Base_JSON/prefixS/json/suki/2022.suki-1.4.json

{
    "paper_id": "2022",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T15:33:34.647240Z"
    },
    "title": "Learning Open Domain Multi-hop Search Using Reinforcement Learning",
    "authors": [
        {
            "first": "Enrique",
            "middle": [],
            "last": "Noriega-Atala",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "The University of Arizona",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Mihai",
            "middle": [],
            "last": "Surdeanu",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "The University of Arizona",
                "location": {}
            },
            "email": "[email protected]"
        },
        {
            "first": "Clayton",
            "middle": [
                "T"
            ],
            "last": "Morrison",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "The University of Arizona",
                "location": {}
            },
            "email": "[email protected]"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "We propose a method to teach an automated agent to learn how to search for multi-hop paths of relations between entities in an open domain. The method learns a policy for directing existing information retrieval and machine reading resources to focus on relevant regions of a corpus. The approach formulates the learning problem as a Markov decision process with a state representation that encodes the dynamics of the search process and a reward structure that minimizes the number of documents that must be processed while still finding multihop paths. We implement the method in an actor-critic reinforcement learning algorithm and evaluate it on a dataset of search problems derived from a subset of English Wikipedia. The algorithm finds a family of policies that succeeds in extracting the desired information while processing fewer documents compared to several baseline heuristic algorithms.",
    "pdf_parse": {
        "paper_id": "2022",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "We propose a method to teach an automated agent to learn how to search for multi-hop paths of relations between entities in an open domain. The method learns a policy for directing existing information retrieval and machine reading resources to focus on relevant regions of a corpus. The approach formulates the learning problem as a Markov decision process with a state representation that encodes the dynamics of the search process and a reward structure that minimizes the number of documents that must be processed while still finding multihop paths. We implement the method in an actor-critic reinforcement learning algorithm and evaluate it on a dataset of search problems derived from a subset of English Wikipedia. The algorithm finds a family of policies that succeeds in extracting the desired information while processing fewer documents compared to several baseline heuristic algorithms.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The sheer size of public corpora such as Wikipedia 1 or large paper repositories like arXiv 2 and PubMed Central 3 poses an enormous challenge to automating effective search for relevant information. This problem is compounded when the underlying information needs require multi-hop connections, e.g., searching for biological mechanisms that connect two proteins (Cohen, 2015) or searching for explanations that require complex reasoning by understanding text supported by different documents in QA systems (Welbl et al., 2018; Yang et al., 2018) .",
                "cite_spans": [
                    {
                        "start": 364,
                        "end": 377,
                        "text": "(Cohen, 2015)",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 508,
                        "end": 528,
                        "text": "(Welbl et al., 2018;",
                        "ref_id": "BIBREF21"
                    },
                    {
                        "start": 529,
                        "end": 547,
                        "text": "Yang et al., 2018)",
                        "ref_id": "BIBREF22"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "In a naive approach, an automated information extraction agent could process all the documents in a corpus, searching for the indirect connections that satisfy a multi-hop information need. However, this quickly becomes prohibitively expensive as the corpus size increases. Further, the documents may also be behind a paywall, adding an additional economic cost to accessing information. Thus, the naive exhaustive reading approach is simply not feasible for most large corpora scenarios. Instead, we need to incorporate the kind of iterative focused reading that humans are capable of. When people search for information, they use background knowledge, based in part on what they have just read, to narrow down the search space while selectively committing time and other resources to carefully reading documents that appear relevant. This process may be repeated multiple times until the information need is satisfied.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "We propose a methodology that uses reinforcement learning (RL) to teach an automated agent how to direct a search process, using existing information retrieval and machine learning components selectively, focusing on the relevant parts of the corpus in order to minimize the expenditure of computational resources and access costs.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The contributions of our work are the following:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "1. A reinforcement learning framework to teach an automated agent how to direct a multi-hop search process that selectively allocates machine reading resources in an open-domain corpus.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "2. A set of domain-agnostic state representation features that enable the reinforcement learning method to learn a policy that improves the chances of finding the desired information while processing fewer documents compared to strong baselines.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "hop search problems derived from English Wikipedia contained in the WikiHop dataset 4 (Welbl et al., 2018) . Using this dataset, we show that our RL approach is able to derive policies that find the desired information more frequently and by processing fewer documents than several heuristic baselines.",
                "cite_spans": [
                    {
                        "start": 86,
                        "end": 106,
                        "text": "(Welbl et al., 2018)",
                        "ref_id": "BIBREF21"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "A new dataset of open-domain multi-",
                "sec_num": "3."
            },
            {
                "text": "Modern machine reading technology enables the extraction of structured information from natural language data. Named-entity recognition (Tjong Kim Sang and De Meulder, 2003) systems detect and label specific classes of concepts from text, both in the general domain (Manning et al., 2014) and for specific domains (Neumann et al., 2019) . Relation extraction systems extract interactions between different concepts in open-domain (DBL, 2018 (DBL, , 2017 (DBL, , 2008 and domain-specific scenarios (Jin-Dong et al., 2019; Demner-Fushman et al., 2019; Cohen et al., 2011) . Reinforcement learning has been successfully deployed for a variety of natural language processing (NLP) tasks. (Clark and Manning, 2016) proposed a policy-gradient method to resolve the correct coreference chains for the task of coreference resolution. (Li et al., 2017) used reinforcement learning to train an end-to-end task-completion dialogue system. For the task of machine translation, (He et al., 2016) formulated the task as a dual-learning game in which two agents teach each other without the need of human labelers using policy-gradient algorithms.",
                "cite_spans": [
                    {
                        "start": 136,
                        "end": 173,
                        "text": "(Tjong Kim Sang and De Meulder, 2003)",
                        "ref_id": "BIBREF19"
                    },
                    {
                        "start": 266,
                        "end": 288,
                        "text": "(Manning et al., 2014)",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 314,
                        "end": 336,
                        "text": "(Neumann et al., 2019)",
                        "ref_id": "BIBREF16"
                    },
                    {
                        "start": 430,
                        "end": 440,
                        "text": "(DBL, 2018",
                        "ref_id": null
                    },
                    {
                        "start": 441,
                        "end": 453,
                        "text": "(DBL, , 2017",
                        "ref_id": null
                    },
                    {
                        "start": 454,
                        "end": 466,
                        "text": "(DBL, , 2008",
                        "ref_id": null
                    },
                    {
                        "start": 497,
                        "end": 520,
                        "text": "(Jin-Dong et al., 2019;",
                        "ref_id": null
                    },
                    {
                        "start": 521,
                        "end": 549,
                        "text": "Demner-Fushman et al., 2019;",
                        "ref_id": null
                    },
                    {
                        "start": 550,
                        "end": 569,
                        "text": "Cohen et al., 2011)",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 684,
                        "end": 709,
                        "text": "(Clark and Manning, 2016)",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 826,
                        "end": 843,
                        "text": "(Li et al., 2017)",
                        "ref_id": "BIBREF13"
                    },
                    {
                        "start": 965,
                        "end": 982,
                        "text": "(He et al., 2016)",
                        "ref_id": "BIBREF9"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "Reinforcement learning has also been specifically applied to improving search and machine reading. In learning how to search, (Kanani and McCallum, 2012) proposed a methodology for the task of slot-filling based on temporal-difference q-learning that uses domain specific state representation features to select actions in a resourceconstrained scenario. (Noriega-Atala et al., 2017) successfully applied RL to finding relevant biochemical interactions in a large corpus by focusing the allocation of machine reading resources towards the most promising documents. Similarly, (Wang et al., 2019) explore the use of deep neural networks and deep RL to simulate the search behavior of a researcher, also in the biomedical domain.",
                "cite_spans": [
                    {
                        "start": 126,
                        "end": 153,
                        "text": "(Kanani and McCallum, 2012)",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 355,
                        "end": 383,
                        "text": "(Noriega-Atala et al., 2017)",
                        "ref_id": "BIBREF17"
                    },
                    {
                        "start": 576,
                        "end": 595,
                        "text": "(Wang et al., 2019)",
                        "ref_id": "BIBREF20"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "We propose a methodology to teach an automated agent how to selectively retrieve and read documents in an iterative fashion in order to efficiently find multi-hop connections between a pair of con-cepts (or entities). Each search step focuses on a restricted set of documents that are hypothesized to be more relevant for finding a connection between the two target concepts. The focus set is retrieved and processed and if a path connecting the concepts is found, the search terminates. Otherwise, a new set of focus documents is identified based on what has been learned so far during the search. The process is repeated iteratively until the desired information is found or a number of iterations is exceeded. Our method is general as it does not directly rely on any supervised domain specific semantics.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "During the search, the agent iteratively constructs a knowledge graph (KG) that represents the relations between concepts found so far through machine reading. In each iteration, the algorithm formulates a document retrieval query based on the current state of the knowledge graph, which is then executed by an information retrieval (IR) component. The IR component contains data structures to query the corpus, for example using an inverted index. The construction of these data structures usually only requires shallow processing, such as tokenization and stemming, and not a full-fledged NLP pipeline. Any documents returned from executing the query are processed by an information extraction (IE) component that performs named entity recognition and relation extraction. The KG is expanded by adding newly identified entities as new nodes and previously unseen relations as new edges. The overall goal of the method is to focus on the documents that appear to be most likely to contain a path between the target concepts, all while processing as few documents as possible.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "(Noriega-Atala et al., 2017) formalized this iterative search process as a family of focused reading algorithms, shown in Algorithm 1:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "Algorithm 1 Focused reading algorithm 1: procedure FOCUSEDREADING(E1, E2) 2: KG \u2190 {{E1, E2}, \u2205} 3: repeat 4: Q \u2190 BUILDQUERY(KG) 5: (V, E) \u2190 RETRIEVAL+EXTRACTION(Q) 6: EXPAND(V, E, KG) 7:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "until ISCONNECTED(E1, E2) OR HASTIMEDOUT 8: end procedure",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "The algorithm starts with the KG representing only the endpoints of the search: the named entities E1 and E2. The algorithm then initiates the search loop. The first step in the loop analyzes the knowledge graph and generates an information retrieval query, Q. As we will describe shortly, the current state of the KG is used to parameterize and constrain the scope of Q, focusing it on returning a limited subset of documents that are hypothesized to be most relevant. After retrieval, the documents are processed by the IE component. Any entities not previously found in the KG are placed in the new entity set V , and similarly any new relations linking entities are placed in the new relation set, E. V and E are incorporated into the KG and the algorithm then searches the updated KG for any new possible paths connecting E1 and E2. If a path exists, it is returned as a candidate explanation of how E1 and E2 are related. Otherwise, if no such path exists, the query formulation process (using the updated KG) followed by IR and IE, is repeated until a path is found or the process times out.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "This framework can answer multi-hop search queries for which the relationships along a connecting path come from different documents. For example, this process may discover that Valley of Mexico is connected to the Aztecs because the Aztecs were a pre-columbian civilization (found in one document), which, in turn, was located in the Valley of Mexico (found in another document).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "In the following subsections, we formulate focused reading as a Markov decision process (MDP). In the focused reading MDP, actions are comprised of information retrieval queries. Actions are constructed from a set of three query templates, listed in Table 1 . Each template is parameterized by one or two arguments representing the entities that are the subject of the query. The template type then incorporates these entities into the set of constraints that must be satisfied by a document in order to be retrieved. The different query templates are intuitively designed to give the agent the choice of either exploring the corpus by performing a broader search through the more permissive disjunctive query (documents are retrieved if either of the entities are present), or instead exploiting particular regions of the corpus through the more restrictive conjunctive query (the documents must contain both entities).",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 250,
                        "end": 257,
                        "text": "Table 1",
                        "ref_id": "TABREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Learning to Search",
                "sec_num": "3"
            },
            {
                "text": "Because conjunctive queries return documents with the text of both entities, they are more likely to identify relations connecting the entities. However, there is also an increased risk that such queries will end up not finding any satisfying documents, especially when the entities are not closely related, resulting in waisting one iteration in the search process. On the other hand, disjunctive queries are designed to return a larger set of documents, which, reduce the likelihood of returning an empty set. But they introduce the risk of processing more potentially irrelevant documents, and potentially introducing more irrelevant entities. Singleton queries represent a compromise between conjunction and disjunction. They are designed to expand the set of existing queries to the knowledge graph, which may in turn be along paths that connect the target entities, but retrieving documents related to just one entity, rather than two.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "Every entity or pair of entities in the current knowledge graph is eligible to serve as a parameter in a query template. The challenge is to choose which entities paired with query template type are more likely to retrieve documents containing candidate paths, using only the domain-agnostic information present in the KG.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "As the search process proceeds, the number of entities in the knowledge graph grows, in turn increasing the number of possible query actions that can be constructed. RL quickly becomes intractable as the state and action space grows. We therefore perform a beam search to fix the cardinality of the action space to a constant size. In particular, we use cosine similarity (for entity pairs) and average tf-idf scores (for single entities) to rank the entities that might participate in constructing query actions. The agent then chooses among the top n entities/pairs for each query template, thus bounding the total number of actions available to the agent to 3n different queries at each step.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "We rank candidate entity pairs that might participate in query templates involving two entities by computing the cosine similarity of the vector representations of the entities. We use the natural language expression representation of the named entities to construct a continuous vector represen-tation. The vector representation of each entity is built by averaging the word embedding vectors 5 of the words in the text of the named entity description. This similarity works as a proxy indicator of how related those entities are, under the intuition that entities that have similar embeddings are more likely to participate in relations.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "For singleton entity queries, we use the average tf-idf score of the entity's natural language description for ranking. The tf-idf score of an entity is derived from averaging the tf-idf score of the individual terms in the entity's natural language description. Each term's frequency value is based on the complete corpus. Tf-idf scores are often used as a proxy measure of term importance (the term occurs selectively with greater frequency within some documents), so here the intuition is that entities with higher tf-idf scores may be associated with higher recall in the corpus.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "Finally, there is an additional non-query action that is available in every step of the search: early stop. If the agent choses to stop early, the search process transitions to a final, unsuccessful state. This deprives the agent from successfully finding a path, but avoids incurring further cost of processing more documents in a possibly unfruitful search.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Constructing Query Actions",
                "sec_num": "3.1"
            },
            {
                "text": "At each step during search, the focused reading agent will select just one action to execute (a query action or early stop) based on the current search state. The agent makes this decision using a model that estimates for each action the expected longterm reward that can be achieved by taking that action in the current state. Here we describe the collection of features used to represent the current state, provided as input to the model. Table 2 provides a summary of the features included in the state representation. We group them into four categories.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 441,
                        "end": 448,
                        "text": "Table 2",
                        "ref_id": "TABREF3"
                    }
                ],
                "eq_spans": [],
                "section": "State Representation Features",
                "sec_num": "3.2"
            },
            {
                "text": "\u2022 Search state features: Information about the current state of the search process including: the number of documents that have been processed so far; how long has the search been running (expressed in iterations); and the size of the knowledge graph.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "State Representation Features",
                "sec_num": "3.2"
            },
            {
                "text": "\u2022 Endpoints of the search: E1 and E2 represent the original target concepts that we are trying 5 We used the pretrained GloVe model provided by spaCy at https://spacy.io/models/en#en_core_web_ lg to find a path between. The identity of the endpoints determines the starting point of the search and conditions the theme of the content sought during the search. This information is provided to the model using the vector embedding representations of E1 and E2.",
                "cite_spans": [
                    {
                        "start": 95,
                        "end": 96,
                        "text": "5",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "State Representation Features",
                "sec_num": "3.2"
            },
            {
                "text": "\u2022 Query features: We include in the state representation features the score with which each of the 3n queries in the action space is ranked. This includes the cosine similarity for conjunction and disjunction queries and tf-idf score for singleton queries. The intuition is that the score may be correlated with the expected long-term reward. For each query action, we will also see the identities of which documents will be retrieved. This allows us to count how many documents will be retrieved that have not already contributed to the KG, and this is included in the state representation for each action.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "State Representation Features",
                "sec_num": "3.2"
            },
            {
                "text": "\u2022 Topic modeling features: Finally, we would like to incorporate some indication of what information is contained in the documents that might be retrieved, and how it relates to the current entities within the KG. As a proxy for this information, we model the topics in the potentially retrieved documents, by peeking into the IR component to see the identity of documents that would be returned by the queries in the action space, and compare them to the topics represented in the KG using two numerical scores: (a) an approximation of how broad or specific the topics are in the set of documents that would be returned by the query, and (b) an estimate of how the knowledge graph's topic distribution would change if the query is selected as the next action.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "State Representation Features",
                "sec_num": "3.2"
            },
            {
                "text": "Topic modeling features can be useful for staying on topic throughout the search, avoiding drift into potentially irrelevant content. Consider the following example: A user wants to know how the Red Sox and the Golden State Warriors are related by searching Wikipedia. While the two entities cover different sports in different regions of the United States, it is more likely that the connection will occur in a document about sports, e.g., they are both covered by the ESPN TV station. We use Latent Dirichlet Allocation (Blei et al., 2003) (LDA) to provide the agent the ability to exploit topic information available in the corpus. LDA is unsupervised and requires only shallow processing of the corpus, namely, tokenizing and optionally stemming. This is essentially the same information required for constructing an inverted index for IR, so can be computed along with the IR component used in the focused reading system.",
                "cite_spans": [
                    {
                        "start": 522,
                        "end": 541,
                        "text": "(Blei et al., 2003)",
                        "ref_id": "BIBREF3"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Topic Modeling Features",
                "sec_num": "3.3"
            },
            {
                "text": "LDA produces a topic distribution for each document. We then aggregate the set of documents by summing the topic frequencies across documents and renormalizing. The topic distribution of the KG is then the aggregation of the topic distributions of the documents processed so far in the search process. The topic distribution of a query is the aggregation of the topic distributions of the unseen documents returned by the query.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Topic Modeling Features",
                "sec_num": "3.3"
            },
            {
                "text": "We consider two statistics for relating topic distributions: topic entropy and Kullback-Leibler (KL) divergence.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Topic Modeling Features",
                "sec_num": "3.3"
            },
            {
                "text": "Intuitively, the entropy of a topic distribution is an estimate of how specialized a document is, that is, how much it focuses on a particular set of topics. For example, a document that only talks about a specific sport will generally have a topic distribution where the mass is concentrated only on the particular topics of that sport, and therefore have a lower entropy than another document that discusses sports and business. Document sets with overall higher entropy are more likely to introduce information about more topics to the knowledge graph, and therefore produce more opportunities for new links between a broader set of entities. Lower entropy queries focus on a narrower set of topics, and thus, may introduce links between a restricted set of entities. The difference in entropy expresses this intuition in relative terms. We introduce a feature, \u2206 Entropy, as the difference in entropy between documents retrieved by a candidate action and the documents the action retrieved in the previous step. Positive values indicate that the candidate query will generally expand the topics compared to those fetched the last step while negative values indicate more restricted topic focus.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Topic Modeling Features",
                "sec_num": "3.3"
            },
            {
                "text": "\u2206 Entropy measures how concentrated the mass is, but it does not tell us how the distributions are different. Two document sets may have completely different topic distributions, yet have the same or similar entropy. Kullback-Leibler (KL) divergence (Kullback and Leibler, 1951) , also known as relative entropy, helps measure how different two distributions are with respect to each other, even if they have the same absolute entropy. To capture this information, we compute the KL divergence between the topic distribution in the new documents (retrieved by the new query) and the topic distribution of the knowledge graph. This estimates how different the information in the new query is relative to what has already been retrieved.",
                "cite_spans": [
                    {
                        "start": 250,
                        "end": 278,
                        "text": "(Kullback and Leibler, 1951)",
                        "ref_id": "BIBREF12"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Topic Modeling Features",
                "sec_num": "3.3"
            },
            {
                "text": "The overall goal of the focused reading learning processes is to identify a policy that efficiently finds paths of relations between the target entities while minimizing the number of documents that must be processed by the IE component. To achieve this, we want the reward structure of the MDP to incorporate the tradeoff between the number of documents that have to be read (the reading cost) and whether the agent can successfully find a path between the entities. Equation 1 describes the reward function, where s t represents the current state and a t represents the action executed in that state.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reward Function Structure",
                "sec_num": "3.4"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "r(st, at) = \uf8f1 \uf8f4 \uf8f2 \uf8f4 \uf8f3 S if st+1 is succesful sate \u2212c \u00d7 m if m > 0 \u2212e if m = 0",
                        "eq_num": "(1)"
                    }
                ],
                "section": "Reward Function Structure",
                "sec_num": "3.4"
            },
            {
                "text": "A positive reward S (for \"success\") is given when executing a t results in a transition to a state whose knowledge graph contains a path connecting the target entities. Otherwise, the search is not yet complete and the a cost is incurred for processing m documents with machine reading on step t. The cost is adjusted by a hyper parameter c that controls the relative expense of processing a single document. Note that there may be actions that return an empty document set, incurring no cost from reading, but still not making progress in the search. To discourage the agent from choosing such actions, the hyperparameter e controls the cost of executing an unfruitful action that returns no new information. (Specific parameter values used in this work are presented in Table 4 of Section 5.)",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 772,
                        "end": 779,
                        "text": "Table 4",
                        "ref_id": "TABREF6"
                    }
                ],
                "eq_spans": [],
                "section": "Reward Function Structure",
                "sec_num": "3.4"
            },
            {
                "text": "To evaluate the focused reading learning method, we introduce a novel dataset derived from the English version of Wikipedia. Our dataset consist of a set of 369 multi-hop search problems, where a search problem is consists of a pair of entities to be connected by a path of relations, potentially connecting to other entities along the path. The foundation of the dataset is a subset of 6,880 Wikipedia articles from the WikiHop (Welbl et al., 2018) corpus. We used Wikification (Ratinov et al., 2011; Cheng and Roth, 2013) to extract named entities from these documents and normalize them to the title of a corresponding Wikipedia article. Wikification does not perform relation extraction, so we lack gold-standard relations. To overcome this limitation, in this paper we induce a relation between entities that co-occur within a window of three sentences. Every relation extracted this way can be traced back to at least one document in the corpus.",
                "cite_spans": [
                    {
                        "start": 429,
                        "end": 449,
                        "text": "(Welbl et al., 2018)",
                        "ref_id": "BIBREF21"
                    },
                    {
                        "start": 479,
                        "end": 501,
                        "text": "(Ratinov et al., 2011;",
                        "ref_id": "BIBREF18"
                    },
                    {
                        "start": 502,
                        "end": 523,
                        "text": "Cheng and Roth, 2013)",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "We create a gold-standard knowledge graph using the induced entities and relations, and we sample pairs of entities connected by paths to create search problems for the dataset. We train an LDA model 6 and constructed an information retrieval inverted index over the collection of documents. 7 We used the Advantage Actor Critic algorithm (Mnih et al., 2016 ) (A2C) to implement our reinforcement learning focused reading method. 8 A2C is an actor-critic method and we use a single neural network architecture to model the action policy (actor) as well as the state value function (critic). We use a single neural network architecture to implement the A2C actor-critic model. The architecture consists of a fully-connected feed-forward neural network with four layers and two output heads. The first output head represents the approximation of the action policy (the actor) as a soft-max activation layer whose size is the cardinality of the action space. This approximates the probability distribution of the actions given the current state. The second head approximates the state value, as a single neuron with a linear activation. The state value estimates the expected long-term reward of using the estimated action distribution of the first head in the current state. Altogether the model consists of approximately 3.79 million parameters. We performed an ablation analysis on the development dataset to find the best configuration of features.",
                "cite_spans": [
                    {
                        "start": 292,
                        "end": 293,
                        "text": "7",
                        "ref_id": null
                    },
                    {
                        "start": 339,
                        "end": 357,
                        "text": "(Mnih et al., 2016",
                        "ref_id": "BIBREF15"
                    },
                    {
                        "start": 430,
                        "end": 431,
                        "text": "8",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "The development dataset contains five hundred search problems. The set of endpoints of the search problems does not overlap with those of the training and validation datasets. This is enforced to avoid any accidental leak of training information. Table 5 contains the results of the ablation experiments. All the search problems were repeated five times with different random seeds. The key columns of the table are defined as follows. Success Rate represents the percentage of problems in the test set for which the agent connected the endpoints. Processed Docs provides the number of documents processed in all the search problems of the test set. Docs per Success is a summary the other two columns: it contains the number of documents processed divided by the number of successes. This ratio is an aggregate statistic useful for comparing the performance between different policies. We report the sample averages and their standard deviations in parentheses. For example, the Success Rate column displays the average and standard deviation of five success rate calculations over five hundred search problems. The Processed Documents column displays the average and standard deviation of the cumulative count of documents processed in the search problems, and so forth. We implement three baseline policies that were not derived using RL:",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 247,
                        "end": 254,
                        "text": "Table 5",
                        "ref_id": "TABREF9"
                    }
                ],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "\u2022 Random: Uniformly randomly selects a query from all possible queries constructed from eligible combinations of entities assigned to the query templates.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "\u2022 Conditional Random: Uniformly randomly selects a query template, conjunction, disjunction and singleton and then choses the uniformly randomly selects the entities to param-eterize the template.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "\u2022 Cascade: Uniformly randomly samples a pair of entities and executes a conjunction query. If the result set does not contain any documents, then the agent selects a disjunction query with the same entities.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "For consistency, each baseline was also evaluated with five different random seeds over the testing set. The top part of Table 5 shows the results of the baseline policies.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 121,
                        "end": 128,
                        "text": "Table 5",
                        "ref_id": "TABREF9"
                    }
                ],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "To test for statistical significance, we performed a non-parametric bootstrap resampling test with ten thousand samples for the the following metrics: success rate, processed documents and documents per success. For each metric, we calculated the difference between the result of the cascade baseline and the result of each of the reinforcement learning (RL) policies. If p \u2264 0.05 of the difference being in favor of the reinforcement learning policy, the quantity is starred in the table.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "In terms of success rate, most of the reinforcement learning models perform better than the cascade baseline. The notable exceptions are two feature configurations that do not use endpoint embeddings. These configurations are the one that considers all feature classes and the one that does not consider topic features. Excluding query features from training produces models that process fewer documents per success with or without endpoint embeddings.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "Excluding search features produced in average models with higher success rate, with or without embeddings, but does so while processing more documents compared to other configurations.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "The configurations that exclude query features produced models with the best numbers of documents per success. When the topic features are excluded, a similar result is achieved, but the number of documents per success of model that does not use endpoint embeddings is not statistically significantly lower than that of the cascade baseline.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "Nonetheless the model that has the best balance between success rate and documents per success is the one that excludes topic features and trains with a dropout coefficient of 0.2 on the endpoint embeddings. We use this model to evaluate the validation dataset. Table 6 displays the results of the cascade baseline, which shows the strongest performance among the baseline policies, and the chosen reinforcement learning model on the validation dataset. The validation dataset contains 650 search problems and the set of endpoints of its search problems is disjoint from the other datasets' for the same reason, to avoid leaking any training or development signal into the validation dataset. We did the same non-parametric bootstrap test for statistical significance. The reinforcement learning policy achieves approximately 2.5% higher average success rate on the testing dataset that the cascade baseline policy. While it also processes fewer documents in average, the difference is not statistically significant, but when considering the number of documents per success, the result is indeed significant, requiring approximately thirty five documents in average than cascade.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 262,
                        "end": 269,
                        "text": "Table 6",
                        "ref_id": "TABREF11"
                    }
                ],
                "eq_spans": [],
                "section": "Evaluation and Discussion",
                "sec_num": "4"
            },
            {
                "text": "We proposed a focused reading methodology to automatically learn how to direct search in large corpora while iteratively building a knowledge base. The knowledge base is modeled as a graph, which in turn is used to focus the search toward documents that appear relevant. Our methodology complements existing information retrieval and machine tools. We evaluated focused reading on a set of search problems extracted from English Wikipedia and demonstrated that reinforcement learning with a state representation based on features about dynamics of the search process and the properties of the corpus is more effective and efficient than heuristic baselines. In this methodology, inference in a knowledge graph acquired during the search process is agnostic of the semantics of the concepts and their relations. Their quality depends on the machine reading components used to extract them.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5"
            },
            {
                "text": "In future work, we plan to explore approaches for incorporating the semantics of the relations along the multi-hop paths that connect the target entities. Crucially, this includes incorporating additional constraints based on topic context. Providing context to a search problem could prove useful to better focus the search process and to improve the accuracy of inference. We also plan to adapt the focused reading methodology to handle other class of search problems, e.g., slot filling tasks where the endpoints are underspecified.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5"
            },
            {
                "text": "https://www.wikipedia.org/ 2 http://arxiv.org/ 3 https://www.ncbi.nlm.nih.gov/pmc/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "http://qangaroo.cs.ucl.ac.uk",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "We used the LDA implementation provided by gensim https://radimrehurek.com/gensim_3.8.3/.7 Code and dataset files are found at https://ml4ai. github.io/OpenDomainFR/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Implemented using the rlpyt library hosted at https: //github.com/astooke/rlpyt.9 Hyper-parameter values were determined through manual tunning.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Proceedings of the First Text Analysis Conference",
                "authors": [],
                "year": 2008,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Proceedings of the First Text Analysis Conference, TAC 2008, Gaithersburg, Maryland, USA, November 17-19, 2008. NIST.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Proceedings of the 2017 Text Analysis Conference",
                "authors": [],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "2017. Proceedings of the 2017 Text Analysis Con- ference, TAC 2017, Gaithersburg, Maryland, USA, November 13-14, 2017. NIST.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Proceedings of the 2018 Text Analysis Conference",
                "authors": [],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "2018. Proceedings of the 2018 Text Analysis Con- ference, TAC 2018, Gaithersburg, Maryland, USA, November 13-14, 2018. NIST.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Latent dirichlet allocation",
                "authors": [
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Blei",
                        "suffix": ""
                    },
                    {
                        "first": "Andrew",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [
                            "I"
                        ],
                        "last": "Jordan",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "Journal of Machine Learning Research",
                "volume": "3",
                "issue": "",
                "pages": "993--1022",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "David Blei, Andrew Ng, and Michael I. Jordan. 2003. Latent dirichlet allocation. Journal of Machine Learning Research, 3:993-1022.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Relational inference for wikification",
                "authors": [
                    {
                        "first": "Xiao",
                        "middle": [],
                        "last": "Cheng",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Roth",
                        "suffix": ""
                    }
                ],
                "year": 2013,
                "venue": "Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "1787--1796",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Xiao Cheng and Dan Roth. 2013. Relational inference for wikification. In Proceedings of the 2013 Con- ference on Empirical Methods in Natural Language Processing, pages 1787-1796, Seattle, Washington, USA. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Improving coreference resolution by learning entitylevel distributed representations",
                "authors": [
                    {
                        "first": "Kevin",
                        "middle": [],
                        "last": "Clark",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Christopher",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Manning",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "643--653",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/P16-1061"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Kevin Clark and Christopher D. Manning. 2016. Im- proving coreference resolution by learning entity- level distributed representations. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 643-653, Berlin, Germany. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Proceedings of BioNLP 2011 Workshop. Association for Computational Linguistics",
                "authors": [
                    {
                        "first": "Kevin",
                        "middle": [],
                        "last": "Bretonnel Cohen",
                        "suffix": ""
                    },
                    {
                        "first": "Dina",
                        "middle": [],
                        "last": "Demner-Fushman",
                        "suffix": ""
                    },
                    {
                        "first": "Sophia",
                        "middle": [],
                        "last": "Ananiadou",
                        "suffix": ""
                    },
                    {
                        "first": "John",
                        "middle": [],
                        "last": "Pestian",
                        "suffix": ""
                    }
                ],
                "year": 2011,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kevin Bretonnel Cohen, Dina Demner-Fushman, Sophia Ananiadou, John Pestian, Jun'ichi Tsujii, and Bonnie Webber, editors. 2011. Proceedings of BioNLP 2011 Workshop. Association for Computa- tional Linguistics, Portland, Oregon, USA.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Darpa's big mechanism program",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "Paul R Cohen",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Physical biology",
                "volume": "12",
                "issue": "4",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Paul R Cohen. 2015. Darpa's big mechanism program. Physical biology, 12(4):045008.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Proceedings of the 18th BioNLP Workshop and Shared Task. Association for Computational Linguistics",
                "authors": [],
                "year": 2019,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Dina Demner-Fushman, Kevin Bretonnel Cohen, Sophia Ananiadou, and Junichi Tsujii, editors. 2019. Proceedings of the 18th BioNLP Workshop and Shared Task. Association for Computational Linguis- tics, Florence, Italy.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Dual learning for machine translation",
                "authors": [
                    {
                        "first": "Di",
                        "middle": [],
                        "last": "He",
                        "suffix": ""
                    },
                    {
                        "first": "Yingce",
                        "middle": [],
                        "last": "Xia",
                        "suffix": ""
                    },
                    {
                        "first": "Tao",
                        "middle": [],
                        "last": "Qin",
                        "suffix": ""
                    },
                    {
                        "first": "Liwei",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Nenghai",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    },
                    {
                        "first": "Tie-Yan",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Wei-Ying",
                        "middle": [],
                        "last": "Ma",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "Advances in neural information processing systems",
                "volume": "",
                "issue": "",
                "pages": "820--828",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Di He, Yingce Xia, Tao Qin, Liwei Wang, Nenghai Yu, Tie-Yan Liu, and Wei-Ying Ma. 2016. Dual learning for machine translation. In Advances in neural information processing systems, pages 820- 828.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Proceedings of The 5th Workshop on BioNLP Open Shared Tasks. Association for Computational Linguistics",
                "authors": [],
                "year": 2019,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kim Jin-Dong, N\u00e9dellec Claire, Bossy Robert, and Del\u00e9ger Louise, editors. 2019. Proceedings of The 5th Workshop on BioNLP Open Shared Tasks. Asso- ciation for Computational Linguistics, Hong Kong, China.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Selecting actions for resource-bounded information extraction using reinforcement learning",
                "authors": [
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Pallika",
                        "suffix": ""
                    },
                    {
                        "first": "Andrew",
                        "middle": [
                            "K"
                        ],
                        "last": "Kanani",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mccallum",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "Proceedings of the fifth ACM international conference on Web search and data mining",
                "volume": "",
                "issue": "",
                "pages": "253--262",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Pallika H. Kanani and Andrew K. McCallum. 2012. Selecting actions for resource-bounded information extraction using reinforcement learning. In Proceed- ings of the fifth ACM international conference on Web search and data mining, pages 253-262.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "On information and sufficiency. The annals of mathematical statistics",
                "authors": [
                    {
                        "first": "Solomon",
                        "middle": [],
                        "last": "Kullback",
                        "suffix": ""
                    },
                    {
                        "first": "Richard",
                        "middle": [
                            "A"
                        ],
                        "last": "Leibler",
                        "suffix": ""
                    }
                ],
                "year": 1951,
                "venue": "",
                "volume": "22",
                "issue": "",
                "pages": "79--86",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Solomon Kullback and Richard A. Leibler. 1951. On information and sufficiency. The annals of mathe- matical statistics, 22(1):79-86.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "End-to-end taskcompletion neural dialogue systems",
                "authors": [
                    {
                        "first": "Xiujun",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Yun-Nung",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Lihong",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Jianfeng",
                        "middle": [],
                        "last": "Gao",
                        "suffix": ""
                    },
                    {
                        "first": "Asli",
                        "middle": [],
                        "last": "Celikyilmaz",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the Eighth International Joint Conference on Natural Language Processing",
                "volume": "1",
                "issue": "",
                "pages": "733--743",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Xiujun Li, Yun-Nung Chen, Lihong Li, Jianfeng Gao, and Asli Celikyilmaz. 2017. End-to-end task- completion neural dialogue systems. In Proceedings of the Eighth International Joint Conference on Nat- ural Language Processing (Volume 1: Long Papers), pages 733-743, Taipei, Taiwan. Asian Federation of Natural Language Processing.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "The Stanford CoreNLP natural language processing toolkit",
                "authors": [
                    {
                        "first": "Christopher",
                        "middle": [
                            "D"
                        ],
                        "last": "Manning",
                        "suffix": ""
                    },
                    {
                        "first": "Mihai",
                        "middle": [],
                        "last": "Surdeanu",
                        "suffix": ""
                    },
                    {
                        "first": "John",
                        "middle": [],
                        "last": "Bauer",
                        "suffix": ""
                    },
                    {
                        "first": "Jenny",
                        "middle": [],
                        "last": "Finkel",
                        "suffix": ""
                    },
                    {
                        "first": "Steven",
                        "middle": [
                            "J"
                        ],
                        "last": "Bethard",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Mc-Closky",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Association for Computational Linguistics (ACL) System Demonstrations",
                "volume": "",
                "issue": "",
                "pages": "55--60",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Christopher D. Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven J. Bethard, and David Mc- Closky. 2014. The Stanford CoreNLP natural lan- guage processing toolkit. In Association for Compu- tational Linguistics (ACL) System Demonstrations, pages 55-60.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Asynchronous methods for deep reinforcement learning",
                "authors": [
                    {
                        "first": "Volodymyr",
                        "middle": [],
                        "last": "Mnih",
                        "suffix": ""
                    },
                    {
                        "first": "Adri\u00e0",
                        "middle": [
                            "Puigdom\u00e8nech"
                        ],
                        "last": "Badia",
                        "suffix": ""
                    },
                    {
                        "first": "Mehdi",
                        "middle": [],
                        "last": "Mirza",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Graves",
                        "suffix": ""
                    },
                    {
                        "first": "Timothy",
                        "middle": [
                            "P"
                        ],
                        "last": "Lillicrap",
                        "suffix": ""
                    },
                    {
                        "first": "Tim",
                        "middle": [],
                        "last": "Harley",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Silver",
                        "suffix": ""
                    },
                    {
                        "first": "Koray",
                        "middle": [],
                        "last": "Kavukcuoglu",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "International conference on machine learning",
                "volume": "",
                "issue": "",
                "pages": "1928--1937",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Volodymyr Mnih, Adri\u00e0 Puigdom\u00e8nech Badia, Mehdi Mirza, Alex Graves, Timothy P. Lillicrap, Tim Harley, David Silver, and Koray Kavukcuoglu. 2016. Asyn- chronous methods for deep reinforcement learning. In International conference on machine learning, pages 1928-1937.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "ScispaCy: Fast and robust models for biomedical natural language processing",
                "authors": [
                    {
                        "first": "Mark",
                        "middle": [],
                        "last": "Neumann",
                        "suffix": ""
                    },
                    {
                        "first": "Daniel",
                        "middle": [],
                        "last": "King",
                        "suffix": ""
                    },
                    {
                        "first": "Iz",
                        "middle": [],
                        "last": "Beltagy",
                        "suffix": ""
                    },
                    {
                        "first": "Waleed",
                        "middle": [],
                        "last": "Ammar",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "Proceedings of the 18th BioNLP Workshop and Shared Task",
                "volume": "",
                "issue": "",
                "pages": "319--327",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/W19-5034"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Mark Neumann, Daniel King, Iz Beltagy, and Waleed Ammar. 2019. ScispaCy: Fast and robust models for biomedical natural language processing. In Pro- ceedings of the 18th BioNLP Workshop and Shared Task, pages 319-327, Florence, Italy. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Learning what to read: Focused machine reading",
                "authors": [
                    {
                        "first": "Enrique",
                        "middle": [],
                        "last": "Noriega-Atala",
                        "suffix": ""
                    },
                    {
                        "first": "Marco",
                        "middle": [
                            "A"
                        ],
                        "last": "Valenzuela-Esc\u00e1rcega",
                        "suffix": ""
                    },
                    {
                        "first": "Clayton",
                        "middle": [],
                        "last": "Morrison",
                        "suffix": ""
                    },
                    {
                        "first": "Mihai",
                        "middle": [],
                        "last": "Surdeanu",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "2905--2910",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/D17-1313"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Enrique Noriega-Atala, Marco A. Valenzuela- Esc\u00e1rcega, Clayton Morrison, and Mihai Surdeanu. 2017. Learning what to read: Focused machine reading. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Pro- cessing, pages 2905-2910, Copenhagen, Denmark. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Local and Global Algorithms for Disambiguation to Wikipedia",
                "authors": [
                    {
                        "first": "Lev",
                        "middle": [],
                        "last": "Ratinov",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Roth",
                        "suffix": ""
                    },
                    {
                        "first": "Doug",
                        "middle": [],
                        "last": "Downey",
                        "suffix": ""
                    }
                ],
                "year": 2011,
                "venue": "Proc. of the Annual Meeting of the Association for Computational Linguistics (ACL)",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lev Ratinov, Dan Roth, Doug Downey, and Mike An- derson. 2011. Local and Global Algorithms for Dis- ambiguation to Wikipedia. In Proc. of the Annual Meeting of the Association for Computational Lin- guistics (ACL).",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition",
                "authors": [
                    {
                        "first": "Erik",
                        "middle": [
                            "F"
                        ],
                        "last": "Tjong",
                        "suffix": ""
                    },
                    {
                        "first": "Kim",
                        "middle": [],
                        "last": "Sang",
                        "suffix": ""
                    },
                    {
                        "first": "Fien",
                        "middle": [],
                        "last": "De Meulder",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "Proceedings of the Seventh Conference on Natural Language Learning at HLT-NAACL 2003",
                "volume": "",
                "issue": "",
                "pages": "142--147",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Erik F. Tjong Kim Sang and Fien De Meulder. 2003. Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition. In Proceedings of the Seventh Conference on Natural Language Learning at HLT-NAACL 2003, pages 142- 147.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "Automatic human-like mining and constructing reliable genetic association database with deep reinforcement learning",
                "authors": [
                    {
                        "first": "Haohan",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Xiang",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Yifeng",
                        "middle": [],
                        "last": "Tao",
                        "suffix": ""
                    },
                    {
                        "first": "Wenting",
                        "middle": [],
                        "last": "Ye",
                        "suffix": ""
                    },
                    {
                        "first": "Qiao",
                        "middle": [],
                        "last": "Jin",
                        "suffix": ""
                    },
                    {
                        "first": "William",
                        "middle": [
                            "W"
                        ],
                        "last": "Cohen",
                        "suffix": ""
                    },
                    {
                        "first": "Eic",
                        "middle": [
                            "P"
                        ],
                        "last": "Xing",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "PSB",
                "volume": "",
                "issue": "",
                "pages": "112--123",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Haohan Wang, Xiang Liu, Yifeng Tao, Wenting Ye, Qiao Jin, William W. Cohen, and Eic P. Xing. 2019. Automatic human-like mining and constructing re- liable genetic association database with deep rein- forcement learning. In PSB, pages 112-123. World Scientific.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Constructing datasets for multi-hop reading comprehension across documents",
                "authors": [
                    {
                        "first": "Johannes",
                        "middle": [],
                        "last": "Welbl",
                        "suffix": ""
                    },
                    {
                        "first": "Pontus",
                        "middle": [],
                        "last": "Stenetorp",
                        "suffix": ""
                    },
                    {
                        "first": "Sebastian",
                        "middle": [],
                        "last": "Riedel",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Transactions of the Association for Computational Linguistics",
                "volume": "6",
                "issue": "",
                "pages": "287--302",
                "other_ids": {
                    "DOI": [
                        "10.1162/tacl_a_00021"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Johannes Welbl, Pontus Stenetorp, and Sebastian Riedel. 2018. Constructing datasets for multi-hop reading comprehension across documents. Transactions of the Association for Computational Linguistics, 6:287- 302.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "HotpotQA: A dataset for diverse, explainable multi-hop question answering",
                "authors": [
                    {
                        "first": "Zhilin",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    },
                    {
                        "first": "Peng",
                        "middle": [],
                        "last": "Qi",
                        "suffix": ""
                    },
                    {
                        "first": "Saizheng",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    },
                    {
                        "first": "William",
                        "middle": [],
                        "last": "Cohen",
                        "suffix": ""
                    },
                    {
                        "first": "Ruslan",
                        "middle": [],
                        "last": "Salakhutdinov",
                        "suffix": ""
                    },
                    {
                        "first": "Christopher",
                        "middle": [
                            "D"
                        ],
                        "last": "Manning",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "2369--2380",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/D18-1259"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christo- pher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 Conference on Empiri- cal Methods in Natural Language Processing, pages 2369-2380, Brussels, Belgium. Association for Com- putational Linguistics.",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF1": {
                "text": "",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF3": {
                "text": "State representation features",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF4": {
                "text": "contains a break-down of the number of elements in each subset of the dataset.",
                "html": null,
                "type_str": "table",
                "content": "<table><tr><td>Element</td><td>Size</td></tr><tr><td>Corpus</td><td>6880 articles</td></tr><tr><td colspan=\"2\">Search Problems Training 230 problems Development 500 problems Testing 670 problems</td></tr><tr><td>Total</td><td>1400 problems</td></tr></table>",
                "num": null
            },
            "TABREF5": {
                "text": "Multi-hop search dataset details.",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF6": {
                "text": "",
                "html": null,
                "type_str": "table",
                "content": "<table><tr><td colspan=\"2\">lists the hyper-parameter values used in our experiments. 9</td></tr><tr><td>Hyper-parameter</td><td>Value</td></tr><tr><td>Environment # entities per query template Maximum # of steps</td><td>15 10</td></tr><tr><td>Reward Function Successful outcome S Document processing cost c Empty query cost e</td><td>1000 10 100</td></tr><tr><td>Training Mini-batch size # Iterations</td><td>100 2000</td></tr></table>",
                "num": null
            },
            "TABREF7": {
                "text": "Hyper-parameter values",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF9": {
                "text": "Feature sets ablation results. * denotes the difference w.r.t. the cascade baseline is statistically significant.",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF11": {
                "text": "Results of the best model in the testing dataset. Quantities are averages over five runs with different random seeds and standard deviations are shown in parentheses. * denotes the difference w.r.t. the cascade baseline is statistically significant.",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            }
        }
    }
}