| { | |
| "paper_id": "O02-1002", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:05:51.583801Z" | |
| }, | |
| "title": "Word Sense Disambiguation and Sense-Based NV Event Frame Identifier", | |
| "authors": [ | |
| { | |
| "first": "Jia-Lin", | |
| "middle": [], | |
| "last": "Tsai", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Academia Sinica", | |
| "location": { | |
| "settlement": "Nankang", | |
| "region": "Taipei", | |
| "country": "Taiwan, R.O.C" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Wen-Lian", | |
| "middle": [], | |
| "last": "Hsu", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Academia Sinica", | |
| "location": { | |
| "settlement": "Nankang", | |
| "region": "Taipei", | |
| "country": "Taiwan, R.O.C" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Jeng-Woei", | |
| "middle": [], | |
| "last": "Su", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Academia Sinica", | |
| "location": { | |
| "settlement": "Nankang", | |
| "region": "Taipei", | |
| "country": "Taiwan, R.O.C" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Word sense is ambiguous in natural language processing (NLP). This phenomenon is particularly keen in cases involving noun-verb (NV) w ord-pairs. This paper describes a sense-based noun-verb event frame (NVEF) identifier that can be used to disambiguate word sense in Chinese sentences effectively. A knowledge representation system (the NVEF-KR tree) for the NVEF sense-pair identifier is also proposed. We use the word sense of Hownet, which is a Chinese-English bilingual knowledge-base dictionary. Our experiment showed that the NVEF identifier was able to achieve 74.8% accuracy for the test sentences studied based only on NVEF sense-pair knowledge. By applying the techniques of longest syllabic NVEF-word-pair first and exclusion word checking, the sense accuracy for the same test sentences could be further improved to 93.7%. There were four major reasons for the incorrect cases: (1) lack of a bottom-up tagger, (2) lack of non-NVEF knowledge, (3) inadequate word segmentation, and (4) lack of a multi-NVEF analyzer. If these four problems could be resolved, the accuracy would reach 98.9%. The results of this study indicate that NVEF sense-pair knowledge is effective for word sense disambiguation and is likely to be important for general NLP.", | |
| "pdf_parse": { | |
| "paper_id": "O02-1002", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Word sense is ambiguous in natural language processing (NLP). This phenomenon is particularly keen in cases involving noun-verb (NV) w ord-pairs. This paper describes a sense-based noun-verb event frame (NVEF) identifier that can be used to disambiguate word sense in Chinese sentences effectively. A knowledge representation system (the NVEF-KR tree) for the NVEF sense-pair identifier is also proposed. We use the word sense of Hownet, which is a Chinese-English bilingual knowledge-base dictionary. Our experiment showed that the NVEF identifier was able to achieve 74.8% accuracy for the test sentences studied based only on NVEF sense-pair knowledge. By applying the techniques of longest syllabic NVEF-word-pair first and exclusion word checking, the sense accuracy for the same test sentences could be further improved to 93.7%. There were four major reasons for the incorrect cases: (1) lack of a bottom-up tagger, (2) lack of non-NVEF knowledge, (3) inadequate word segmentation, and (4) lack of a multi-NVEF analyzer. If these four problems could be resolved, the accuracy would reach 98.9%. The results of this study indicate that NVEF sense-pair knowledge is effective for word sense disambiguation and is likely to be important for general NLP.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Word sense disambiguation (WSD) has been a pervasive problem in natural language processing (NLP) since 1949 [Weaver 1949] . Word sense ambiguity (or lexical ambiguity), is generally classified into two types: syntactic and semantic ambiguity [Small et al. 1988 , Krovetz et al. 1992 . Syntactic ambiguity is caused by differences in syntactic categories (e.g.", | |
| "cite_spans": [ | |
| { | |
| "start": 109, | |
| "end": 122, | |
| "text": "[Weaver 1949]", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 243, | |
| "end": 261, | |
| "text": "[Small et al. 1988", | |
| "ref_id": "BIBREF8" | |
| }, | |
| { | |
| "start": 262, | |
| "end": 283, | |
| "text": ", Krovetz et al. 1992", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "\"play\" can occur as a noun or verb). Semantic ambiguity is caused by homonymy (e.g. \"bank\" in \"to put money in a bank,\" \"the bank of a river\") or polysemy (e.g. \"face\" in \"human face,\" \"face of a clock\"). Although many approaches have been adopted to disambiguate word sense, algorithms for word sense determination still are not reliable [Krovetz et al. 1992 , Resnik et al. 2000 . Human beings usually can disambiguate word sense by using additional information from the speaker, the writer or the context. When out-of-context (or out-of-sentence) information is not symbolized and processed in the computer, WSD either becomes very difficult or, sometimes, impossible. Therefore, it is crucial to investigate what kind of knowledge is useful for WSD [Krovetz et al. 1992 ].", | |
| "cite_spans": [ | |
| { | |
| "start": 339, | |
| "end": 359, | |
| "text": "[Krovetz et al. 1992", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 360, | |
| "end": 380, | |
| "text": ", Resnik et al. 2000", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 753, | |
| "end": 773, | |
| "text": "[Krovetz et al. 1992", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "According to a study in cognitive science [Choueka et al. 1983] , people often disambiguate word sense using only a few other words in a given context (frequently only one additional word). Thus, the relationships between one word and others can be effectively used to resolve ambiguity. Furthermore, from [Small et al. 1988 , Krovetz et al. 1992 , Resnik et al. 2000 , most ambiguities occur with nouns and verbs, and the object-event (i.e. noun-verb) distinction is a major ontological division for humans [Carey 1992 ]. However, no clear data has been collected to support these claims. These observations motivated us to demonstrate through an experiment, how noun-verb (NV) relationships can be used to disambiguate word sense in Chinese sentences.", | |
| "cite_spans": [ | |
| { | |
| "start": 42, | |
| "end": 63, | |
| "text": "[Choueka et al. 1983]", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 306, | |
| "end": 324, | |
| "text": "[Small et al. 1988", | |
| "ref_id": "BIBREF8" | |
| }, | |
| { | |
| "start": 325, | |
| "end": 346, | |
| "text": ", Krovetz et al. 1992", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 347, | |
| "end": 367, | |
| "text": ", Resnik et al. 2000", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 508, | |
| "end": 519, | |
| "text": "[Carey 1992", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "In this paper, we shall focus on word sense disambiguation involving NV word-pairs since these are most troublesome. Consider the following sentence: \" (This car moves well).\" In this sentence, we have two possible NV word-pairs, \" -(car, move)\" and \" -(auto-shop, move).\" It is clear that the permissible NV word-pair is \" -(car, move).\" We shall call such a permissible NV word-pair an NV-event frame (NVEF) word-pair. Using a collection of pre-learned NVEF word-pairs, we can identify the NVEF word-pair \" -\" from the sentence \" .\" The word \" \" in a dictionary can have three possible senses: 'surname' (noun), 'car' (noun) and 'turn' (ve rb). To resolve this ambiguity, we can use the pre-defined sense of the NVEF word-pair \" -(car, move)\" to determine that the correct sense of the Chinese word \" \" is \"car\" in the above Chinese sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "In this paper, we shall show that knowledge of NVEF sense-pairs (to be defined in Section 2) can be effectively used to resolve word sense ambiguity. In the next section, we will propose an NVEF sense-pair identifier, which is based on pre-stored knowledge of NVEF sense-pairs. We use this NVEF sense-pair identifier to identify NVEF word-pairs in an input sentence and to determine the corresponding word senses. In Section 3, we will present and analyze the results of a WSD experiment on a set of test sentences using the NVEF sense-pair identifier. Finally, we will give conclusions and directions for future research in Section 4.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "We use Hownet [Dong] as our system's Chinese machine-readable dictionary (MRD). Hownet is a Chinese-English bilingual knowledge-base dictionary, which provides knowledge of the Chinese lexicon, parts-of-speech (POS) and word senses.", | |
| "cite_spans": [ | |
| { | |
| "start": 14, | |
| "end": 20, | |
| "text": "[Dong]", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Development of an NVEF Sense-Pair Identifier", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "The sense of a word is defined as its DEF (concept definition) in Hownet. Table 1 lists three different senses of the Chinese word \" (Che/car/turn).\" In Hownet, the DEF of a word consists of its main feature and secondary features. For example, in the DEF \"character| ,surname| ,human| ,ProperName| \" of the word \" (Che),\" the first item \"character| \" is the main feature, and the remaining three items, \"surname| ,\" \"human| ,\" and \"ProperName| ,\" are its secondary features. The main feature in Hownet can inherit features in the hypernym-hyponym hierarchy. There are approximately 1,500 features in Hownet. Each of these features is called a sememe, which refers to the smallest semantic unit that cannot be further reduced. The Hownet dictionary used in this study contains 50,121 Chinese words, among which there are 29,719 nouns, 16,652 verbs and 16,242 senses (including 9,893 noun-senses and 4,440 verb-senses). Table 2 gives the statistics of the number of senses per Chinese word and the number of Chinese words per sense used in Hownet. Now, take the NV word-pair \" -(car, move)\" for example. According to the sense of the Chinese word \" (Che/car/turn)\" and the sense of the Chinese word \" (move),\" the only permissible NVEF sense-pair for the NV word-pair \" -(car, move)\" is \"LandVehicle| \"-\"VehicleGo| .\" We call such a permissible NV sense-pair an NVEF sense-pair in this paper. Note that an NVEF sense-pair is a class that includes the permissible word-pair instance \" -(car, move).\"", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 74, | |
| "end": 81, | |
| "text": "Table 1", | |
| "ref_id": "TABREF0" | |
| }, | |
| { | |
| "start": 919, | |
| "end": 926, | |
| "text": "Table 2", | |
| "ref_id": "TABREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Definition of an NVEF Sense-Pair", | |
| "sec_num": "2.1" | |
| }, | |
| { | |
| "text": "A knowledge representation tree (KR-tree) of NVEF sense-pairs is shown in Fig.1 . There are two types of nodes in the KR-tree, namely, function nodes and concept nodes. Concept nodes refer to words and features in Hownet. Function nodes are used to define the relationships between their parent and children concept nodes. If a concept node A is the child of another concept node B, then A is a subclass of B. Following this convention, we can omit the function node \"subclass\" (which should exist) between A and B. We can classify the noun-sense class ( ) into 15 subclasses according to their main features. They are \" (bacteria),\" \" (animal),\" \" (human),\" \" (plant),\" \" (artifact),\" \" (natural),\" \" (event),\" \" (mental),\" \" (phenomena),\" \" (shape),\" \" (place),\" \" (location),\" \" (time),\" \" (abstract)\" and \" (quantity).\" Appendix A gives a sample table of 15 main features of nouns in each noun-sense subclass. subclasses, we have designed three subclass sense-symbols, in which \"=\" means \"exact,\" \"&\" means \"like,\" and \"%\" means \"inclusive.\" An example using these symbols is given below.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 74, | |
| "end": 79, | |
| "text": "Fig.1", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "Given three senses S 1 , S 2 and S 3 defined by a main feature A and three secondary features B, C and D, let", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "S 1 = A, B, C, D,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "S 2 = A, B, and", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "S 3 = A, C, D.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "Then, we have that sense S 2 is in the \"=A,B\" exact-subclass; senses S 1 and S 2 are in the \"&A,B\" like-subclass; and senses S 1 S 2 , and S 3 are in the \"%A\"", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "inclusive-subclass.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "(2) Word-Instance ( ): The content of its children are the words belonging to the sense subclass of its parent node. These words are self-learned by the NVEF sense-pair identifier according to the sentences under the Test-Sentence nodes.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "(3) Test-Sentence ( ): The content of its children is several selected test sentences in support of its corresponding NVEF subclass sense-pair.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Representation Tree of NVEF Sense-Pairs", | |
| "sec_num": "2.2" | |
| }, | |
| { | |
| "text": "To speedup the creation of the KR-tree, an example-based algorithm is proposed to generate the KR-tree semi-automatically. This algorithm is described below.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 1. Select a noun-sense, such as \"disease| ,\" in Hownet.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 2. Collect all Chinese polysyllabic words of the selected noun-sense. (Monosyllabic words are not considered at this stage.)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 3. Select those Chinese un-segmented sentences that include at least one word collected in Step 2 from the Sinica corpus (which is a Chinese corpus of two millions words [CKIP 1995]) or other domain specific collections. For example, the Chinese sentence \" (A doctor's job is to prevent a disease and to cure the patient)\" is a candidate sentence that includes the Chinese word \" (disease).\"", | |
| "cite_spans": [ | |
| { | |
| "start": 175, | |
| "end": 187, | |
| "text": "[CKIP 1995])", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 4. Find all possible verb-senses from the sentences selected in Step 3 to form all possible verb-senses for the selected noun-sense. Calculate the frequency for each verb-sense.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 5. Sort all possible different verb-senses according to their corresponding frequencies from large to small. (See Fig. 2 ) Determine a cut-off frequency in the list. Among all verb-senses above the cut-off frequency, manually pick the permissible ones for the selected noun-sense. Meanwhile, determine their subclass sense-symbols (i.e. \"&,\" \"%\" and \"=\".)", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 119, | |
| "end": 125, | |
| "text": "Fig. 2", | |
| "ref_id": "FIGREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Step 6. Add these permissible NVEF subclass sense-pairs to the KR-tree.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Note that among the above steps, only step 5 requires human intervention. This step is quite laborious, but through learning, human involvement can be greatly reduced. Fig. 2 shows the top 5 possible verb-senses picked by the above algorithm for the noun-sense \"disease| \" collected from 302 sentences in the Sinica corpus. In Fig. 2 , the permissible verb-senses for the noun-sense \"disease| \" are \"cure| \" with a frequency of 24, \"Cause Affect| , medical| \" with one of 23, \"Result In| \" with one of 19 and \"obstruct| \" with one of 14. It is observed that, if the number of sentences collected in Step 3 is greater than 300, then the top 5 verb-senses will almost always form NVEF sense-pairs with the selected noun-sense. ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 168, | |
| "end": 174, | |
| "text": "Fig. 2", | |
| "ref_id": "FIGREF1" | |
| }, | |
| { | |
| "start": 324, | |
| "end": 333, | |
| "text": "In Fig. 2", | |
| "ref_id": "FIGREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Generation of NVEF Sense-Pairs", | |
| "sec_num": "2.3" | |
| }, | |
| { | |
| "text": "Based on the KR-tree, we shall develop a primitive NVEF sense-pair identifier as follows. For a given sentence, the algorithm will first identify all NVEF sense-pairs in the KR-tree that have corresponding NVEF word-pairs in the sentence. It will then arrange these NVEF sense-pairs and their corresponding NVEF word-pairs into a tree, called a sentence-NVEF tree, as shown in Fig. 3 . A more formal description of the primitive NVEF sense-pair identifier is given below:", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 377, | |
| "end": 383, | |
| "text": "Fig. 3", | |
| "ref_id": "FIGREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "Step 1. Input a sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "Step 2. Generate all possible NV word-pairs of the input sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "Step 3. Check each NV word-pair got in step 2 to see if its corresponding NV sense-pairs can be matched to an NVEF subclass sense-pair in the KR-tree. If matches are found, then use the corresponding noun-senses and verb-senses to form the permissible NVEF sense-pairs, respectively, for this sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "Step 4. Arrange all permissible NVEF sense-pairs and their corresponding NVEF word-pairs in a sentence-NVEF tree.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "A system overview of the primitive NVEF sense-pair identifier is given in Fig. 4 . ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 74, | |
| "end": 80, | |
| "text": "Fig. 4", | |
| "ref_id": "FIGREF3" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "A Primitive NVEF Sense-Pair Identifier", | |
| "sec_num": "2.4" | |
| }, | |
| { | |
| "text": "In Fig. 3 , the correct segmented results of the two Chinese sentences are \" / / / / \" and \" / / / ,\" respectively. The upper part of Fig. 3 is a sentence-NVEF tree with a single NVEF sense-pair, \"LandVehicle| \"-\"VehicleGo| ,\" which has two corresponding NV word-pairs, i.e. \" -\" and \" -.\" If we further apply the \"longest syllabic NVEF-word-pair first\" strategy (LS-NVWF), the incorrect NVEF word-pair \" -\" will be successfully dropped. Note that the \"longest syllabic word first strategy\" is an effective technique for Chinese word segmentation [Chen et al. 1986] . The lower part of Fig. 3 is a sentence-NVEF tree with two NVEF sense-pairs including \"expel| \"-\"livestock| \" (NV word-pair is \" -\") and \"facilities| , space| , @foster| , #livestock| \"-\"GoInto| \" (NV word-pair is \" -\").", | |
| "cite_spans": [ | |
| { | |
| "start": 547, | |
| "end": 565, | |
| "text": "[Chen et al. 1986]", | |
| "ref_id": "BIBREF1" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 3, | |
| "end": 9, | |
| "text": "Fig. 3", | |
| "ref_id": "FIGREF2" | |
| }, | |
| { | |
| "start": 134, | |
| "end": 140, | |
| "text": "Fig. 3", | |
| "ref_id": "FIGREF2" | |
| }, | |
| { | |
| "start": 586, | |
| "end": 592, | |
| "text": "Fig. 3", | |
| "ref_id": "FIGREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Another useful technique is to exclude certain nouns or verbs from the sentence-NVEF tree. A word with very low frequency as a noun or a verb is treated as a word of exclusion for the NVEF sense-pair identifier. Take the Chinese word \" (of/target)\" as an example. Its frequency as a noun or a verb is only 0.004% (computed according to the Sinica corpus). Thus, \" \" becomes a word of exclusion. In our experiment, the exclusion word list (EWL) consists of those words whose frequencies as nouns or verbs are no greater than 5%. When an NVEF word-pair includes at least one exclusion word, its corresponding NVEF sense-pair is excluded from the sentence-NVEF tree. This process is called EWL checking. Appendix B lists all of the exclusion words used in this experiment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Thus, our final NVEF sense-pair identifier can be described as follows.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Step 1. Input a sentence.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Step 2. Generate all possible NV word-pairs of the input sentence. Exclude certain word-pairs based on EWL checking.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Step 3. Check each NV word-pair to see if its corresponding NV sense-pairs can be matched to an NVEF subclass sense-pair in the KR-tree. For each NV sense-pair that matches an NVEF subclass sense-pair in the KR-tree, use it to the set of permissible NVEF sense-pairs, respectively, for this sentence. Resolve conflicts using the LS-NVWF strategy.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Step 4. Arrange all permissible NVEF sense-pairs and their corresponding NVEF word-pairs in a sentence-NVEF tree.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "A system overview of the NVEF sense-pair identifier is given in Fig. 5 . To evaluate the WSD performance of the NVEF sense-pair identifier, we will consider a WSD experiment in the next section.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 64, | |
| "end": 70, | |
| "text": "Fig. 5", | |
| "ref_id": "FIGREF4" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "An NVEF Sense-Pair Identifier", | |
| "sec_num": "2.5" | |
| }, | |
| { | |
| "text": "Within a sentence, the number of available NVEF sense-pairs is finite. Consider the Chinese sentence \"", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The WSD experiment", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "(This car moves well).\" Table 3 gives eight possible pairs of NVEF senses found in this sentence, but there is only one permissible NVEF sense-pair, \"LandVehicle| \"-\"VehicleGo| .\" characters per sentence (of the 445 Chinese test sentences) were 4, 24 and 11.5, respectively. In addition, the numbers of single-NVEF sentences and multi-NVEF sentences among the test sentences were 96 and 349, respectively.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 24, | |
| "end": 31, | |
| "text": "Table 3", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The WSD experiment", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "We conducted the experiment in a progressive manner. The NVEF sense accuracy of the test sentences determined using the NVEF sense-pair identifier with only the knowledge of the KR-tree was 74.8% (see Table 4 ). When the strategy of adopting the longest syllabic NVEF-word-pair first (LS-NVWF) was used together with the NVEF sense-pair identifier, the NVEF sense accuracy reached 87.6%. When the exclusion word list (EWL checking) was adopted together with the NVEF sense-pair identifier, the NVEF sense accuracy reached 89.2%. When the techniques of both LS-NVWF and EWL checking were adopted with the NVEF sense-pair identifier (see Table 4 ), the NVEF sense accuracy improved to 93.7%. Meanwhile, along with the NVEF sense-pair identifier, the word-segmentation accuracy (for those ambiguous NVEF word-pairs) for these sentences was 99.6% (443/445). This result also supports the aforementioned claim that the NVEF word-segmentation accuracy was better than the NVEF sense accuracy. Appendix C presents two successful and one unsuccessful sentence-NVEF trees obtained in this experiment. a \"Using LS-NVWF\" represents NVEF sense accuracy using LS-NVWF with the NVEF sense-pair identifier. b \"Using EWL\" represents NVEF sense accuracy using EWL checking with the NVEF sense-pair identifier. c \"Using Both\" represents NVEF sense accuracy using both LS-NVWF and EWL checking with the NVEF sense-pair identifier.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 201, | |
| "end": 208, | |
| "text": "Table 4", | |
| "ref_id": "TABREF2" | |
| }, | |
| { | |
| "start": 636, | |
| "end": 643, | |
| "text": "Table 4", | |
| "ref_id": "TABREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The WSD experiment", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Although the NVEF sense accuracy could reach 93.7% when the techniques of both LS-NVWF and EWL checking were adopted with the NVEF sense-pair identifier, there was still a room for improvement. Below, we have classified the reasons behind the unsuccessful cases into four major types:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Analysis of the Unsuccessful Cases", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "(1) Lack of a bottom-up tagger: There are many specific linguistic units, such as names, addresses, determinative -measure compounds, etc. in sentences which need to be recognized in order to supplement the NVEF sense-pair identifier (which works in a top-down fashion).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Analysis of the Unsuccessful Cases", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "In this study, 6 sentences were unsuccessful for this reason. Although the techniques of LS-NVWF and EWL checking inadvertently resolved these cases, this is still a potential problem.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Analysis of the Unsuccessful Cases", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "include the correct NVEF word-pair for word segmentation. However, the converse is not true. That is, a correct NVEF word-pair cannot guarantee that the corresponding NVEF sense-pair is permissible. Thus, the NVEF word-segmentation accuracy is normally better than the NVEF sense accuracy.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Analysis of the Unsuccessful Cases", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "In this paper, we have described an NVFE sense-pair identifier which we attempted to use to disambiguate word sense in Chinese sentences. A WSD experiment was conducted using the NVEF sense-pair identifier with the KR-tree. The knowledge in the KR-tree was created with the help of a semi-automatic NVEF generation tool.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Directions for Future Research", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Based on current techniques, our experiment showed that the NVEF sense accuracy reached 93.7% and the NVEF word-segmentation accuracy 99.6%. We have indicated, in Section 3, several ways to further improve the performance of our system, some of which are currently being studied.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Directions for Future Research", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Our experiment indicated that NVEF sense-pair knowledge can be used effectively to achieve NVEF word-sense disambiguation in Chinese sentences. It also supports the claim in [Choueka et al. 1983 ] that people usually disambiguate word sense using only a few words (frequently only one word) in the given context. We are particularly pleased to note that the NVEF knowledge can achieve high accuracy in NVEF word-segmentation since correct word-segmentation is one key to a successful Chinese NLP [Slocum et al. 1985] .", | |
| "cite_spans": [ | |
| { | |
| "start": 174, | |
| "end": 194, | |
| "text": "[Choueka et al. 1983", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 496, | |
| "end": 516, | |
| "text": "[Slocum et al. 1985]", | |
| "ref_id": "BIBREF7" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Directions for Future Research", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Although we have a semi-automatic NVEF generation tool, it was still a laborious task to create our current level of NVEF knowledge, which constitutes only 7.7% of the entire NVEF knowledge. Hence, a systematic method for fully automatic NVEF knowledge generation is highly desired. Furthermore, we will try to develop a combined top-down and bottom-up NVEF sense-pair identifier that can address the issues involved in the four unsuccessful cases described in Section 3.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Directions for Future Research", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "We plan to create a full fledged KR-tree so that we can investigate the robustness of the sense-based approach for monolingual and bilingual (e.g. English-Chinese) WSD. The study of NVEF will also be extended to noun-noun pairs, noun-adjective pairs and verb-adverb pairs. Another related research goal is to apply the NVEF sense-pair identifier to other fields of NLP, in particular, document classification, information retrieval, question answering and speech understanding.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Directions for Future Research", | |
| "sec_num": "4." | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "We are grateful to the our colleagues in the Intelligent Agent Systems Lab., Li-Yeng Chiu, Mark Shia, Gladys Hsieh, Masia Yu and Yi -Fan Chang, who helped us create all the necessary NVEF knowledge for this study. We would also like to thank Prof. Zhen-Dong", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Acknowledgements", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "To evaluate the performance of WSD by using the NVEF sense-pair identifier with the KR-tee, we define the NVEF sense accuracy for a set of test sentences to be NVEF sense accuracy = # of successful sentences / # of test sentences,(1) where a sentence is successful if all NVEF sense-pairs and their corresponding NVEF word-pairs obtained from the NVEF sense-pair identifier are correct for this sentence. With the KR-tree, the WSD performance for the test sentences can be evaluated by computing the NVEF sense accuracy. This equation is designed from the viewpoint of natural language understanding. Since NVEF sense-pairs often represent a key feature in the meaning of a sentence, any incorrect NVEF sense-pair identification could result in misunderstanding this sentence. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "annex", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The framework of WSD evaluation for the NVEF sense-pair identifier is as follows.1.Select a set of Chinese test sentences from the Sinica Corpus [CKIP 1995] randomly.2.Use the tool of example-based possible NVEF generation to search and create all permissible NVEF subclass sense-pairs found in these test sentences in the KR-tree.3.Apply the NVEF sense-pair identifier to these test sentences and obtain their corresponding sentence-NVEF trees.4. Compute the NVEF sense accuracy for the test sentences using Equation 1.In this study, we analyzed 7.7% (=764/9,893) of the noun-senses in Hownet and created 4,028 NVEF subclass sense-pairs in the KR-tree. The minimum, maximum and mean of (2) Lack of Non-NVEF knowledge: Consider the Chinese sentence, \" (A wife wants to take her husband's wallet into her hands).\" There were three different noun-senses of the Chinese word, \"(teacher/doctor/husband),\" which could form an NVEF sense-pair with the verb-sense \" (take\u2026into one's hands).\" To get the correct noun-sense \" (husband)\" for this sentence, we need the knowledge of a noun-noun (NN) sense-pair, such as \" (wife)\"-to-\" (husband),\" or other contextual information.This knowledge is not available from the KR-tree and needs to be collected separately. In this study, 15 sentences were unsuccessful for this reason, and this problem could not be resolved using the technique of LS-NVWF or EWL checking.(3) Inadequate word segmentation: Consider the Chinese sentence, \" (He obtained the championship with a full mark).\" There were two possible verbs with the same verb-sense \" (obtain)\" and \" (obtain)\" that could form NVEF sense-pairs with the noun-sense \" (champ).\" In this case, we have two conflicting NVEF sense-pairs and need a better segmentation algorithm to determine that the correct verb are \" (obtain)\"for this sentence (the correct segmented result of this sentence is \" / / / / \").In this study, 3 sentences were unsuccessful for this reason, and this problem could not be resolved using the technique of LS-NVWF or EWL checking.(4) Lack of a multi-NVEF analyzer: Consider the Chinese sentence \" (Take airplane to leave Taipei).\" The NVEF sense-pair identifier detected that there were three NVEF sense-pairs: multi-NVEF sense-pairs. In this study, 5 sentences were unsuccessful for this reason, and this problem could not be resolved using the technique of LS-NVWF or EWL checking.If these four problems could be resolved, the NVEF sense accuracy could be improved to (417+15+3+5) / (445) = 98.9%.Based on this experiment, we find that our NVEF sense-pair identifier has the potential to provide the following information for a given sentence: (1) main verbs, (2) nouns, (3) NVEF word-pairs, (4) NVEF sense-pairs, (5) NVEF phrase-boundaries, and (6) the initial relationship among multi-NVEF sense/word-pairs. A correct NVEF sense-pair will naturally ", | |
| "cite_spans": [ | |
| { | |
| "start": 145, | |
| "end": 156, | |
| "text": "[CKIP 1995]", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "WSD Evaluation", | |
| "sec_num": "3.1" | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "The origin and evolution of everyday concepts", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Carey", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "Cognitive Models of Science", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Carey, S., \"The origin and evolution of everyday concepts (In R. N. Giere, ed.),\" Cognitive Models of Science, Minneapolis: University of Minnesota Press, 1992.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "A model for Lexical Analysis and Parsing of Chinese Sentences", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "G" | |
| ], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Chen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "Proceedings of 1986 International Conference on Chinese Computing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "33--40", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chen, C. G., K. J. Chen and L. S. Lee, \"A model for Lexical Analysis and Parsing of Chinese Sentences,\" Proceedings of 1986 International Conference on Chinese Computing, Singapore, 1986, pp.33-40.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A Connectionist Scheme for Modeling Word Sense Disambiguation", | |
| "authors": [ | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Choueka", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Lusignan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1983, | |
| "venue": "Cognition and Brain Theory", | |
| "volume": "6", | |
| "issue": "1", | |
| "pages": "89--120", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Choueka, Y. and S. Lusignan, \"A Connectionist Scheme for Modeling Word Sense Disambiguation,\" Cognition and Brain Theory, 6 (1) 1983, pp.89-120.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "the content and illustration of Sinica corpus of Academia Sinica", | |
| "authors": [], | |
| "year": 1995, | |
| "venue": "CKIP", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "CKIP. Technical Report no. 95-02, the content and illustration of Sinica corpus of Academia Sinica. Institute of Information Science, Academia Sinica, 1995. http://godel.iis.sinica.edu.tw/CKIP/r_content.html", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Lexical Ambiguity and Information Retrieval", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Krovetz", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "B" | |
| ], | |
| "last": "Croft", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "ACM Transactions on Information Systems", | |
| "volume": "10", | |
| "issue": "2", | |
| "pages": "115--141", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Krovetz, R. and W. B. Croft, \"Lexical Ambiguity and Information Retrieval,\" ACM Transactions on Information Systems, 10 (2), 1992, pp.115-141.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Distinguishing Systems and Distinguishing Senses: New Evaluation Methods for Word Sense Disambiguation", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Resnik", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Yarowsky", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Natural Language Engineering", | |
| "volume": "5", | |
| "issue": "3", | |
| "pages": "113--133", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Resnik, P. and D. Yarowsky, \"Distinguishing Systems and Distinguishing Senses: New Evaluation Methods for Word Sense Disambiguation,\" Natural Language Engineering, 5 (3), 2000, pp.113-133.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Transportability to Other Language: The Natural Language Processing Project in the AI Program at MCC", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Slocum", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "F" | |
| ], | |
| "last": "Justus", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1985, | |
| "venue": "ACM Transactions on Office Information Systems", | |
| "volume": "3", | |
| "issue": "2", | |
| "pages": "204--230", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Slocum, J. and C. F. Justus, \"Transportability to Other Language: The Natural Language Processing Project in the AI Program at MCC,\" ACM Transactions on Office Information Systems, 3(2) 1985, pp.204-230.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Lexical Ambiguity Resolution", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Small", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Cottrell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "E" | |
| ], | |
| "last": "Tannenhaus", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Small, S., and G. Cottrell, and M. E. Tannenhaus, Lexical Ambiguity Resolution, Morgan Kaufmann, Palo Alto, Calif., 1988.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Using WordNet to Disambiguate Word Senses for Text Retrieval", | |
| "authors": [ | |
| { | |
| "first": "E", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Voorhees", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "ACM-SIGIR", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "171--180", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Voorhees, E. M., \"Using WordNet to Disambiguate Word Senses for Text Retrieval,\" ACM-SIGIR, 1993, pp. 171-180.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "An illustration of the KR-tree using \" (artifact)\" as an example noun-sense subclass. (The English words in parentheses are there for explanatory purposes only.)Three function nodes are used in the KR-tree as shown inFig. 1:(1) Major-Event ( ): The content of its parent node represents a noun-sense subclass, and the content of its child node represents a verb-sense subclass. A noun-sense subclass and a verb-sense subclass linked by a Major-Event function node is an NVEF subclass sense-pair, such as \"&LandVehicle| \" and \"=VehcileGo| \" inFig. 1. To describe various relationships between noun-sense and verb-sense", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "Top 5 possible verb-senses for creating permissible NVEF sense subclasses for the noun-sense \"disease| .\" Word Sense Disambiguation and Sense-Based NV Event Frame Identifier 35", | |
| "num": null | |
| }, | |
| "FIGREF2": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "Two sentence-NVEF trees for the input Chinese sentences (a) \" \" (a single-NVEF sentence) and (b) \" \" (a multi-NVEF sentence), respectively.", | |
| "num": null | |
| }, | |
| "FIGREF3": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "System overview of the primitive NVEF sense-pair identifier.", | |
| "num": null | |
| }, | |
| "FIGREF4": { | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "A system overview of the NVEF sense-pair identifier.", | |
| "num": null | |
| }, | |
| "TABREF0": { | |
| "num": null, | |
| "html": null, | |
| "type_str": "table", | |
| "content": "<table><tr><td>Che</td><td>Noun</td><td>character|</td><td>, surname| , human| , ProperName|</td></tr><tr><td>car</td><td>Noun</td><td>LandVehicle|</td><td/></tr><tr><td>turn</td><td>Verb</td><td>cut|</td><td/></tr><tr><td colspan=\"4\">a C.Word refers to a Chinese word; E.Word refers to an English word</td></tr></table>", | |
| "text": "Three different senses of the Chinese word \" (Che/car/turn).\" C.Word a E.Word a Part-of-speech Sense (i.e. DEF in Hownet)" | |
| }, | |
| "TABREF1": { | |
| "num": null, | |
| "html": null, | |
| "type_str": "table", | |
| "content": "<table><tr><td>Item a</td><td>Total</td><td>Noun</td><td>Verb</td></tr><tr><td>Maximum number of senses per Chinese word</td><td>27</td><td>14</td><td>24</td></tr><tr><td>Mean number of senses per Chinese word</td><td>1.24</td><td>1.14</td><td>1.23</td></tr><tr><td>Maximum number of Chinese words per sense</td><td>374</td><td>372</td><td>129</td></tr><tr><td>Mean number of Chinese words per sense</td><td>3.8</td><td>3.0</td><td>4.6</td></tr></table>", | |
| "text": "Statistics of the number of senses per Chinese word and the number of Chinese words per sense used in Hownet." | |
| }, | |
| "TABREF2": { | |
| "num": null, | |
| "html": null, | |
| "type_str": "table", | |
| "content": "<table><tr><td colspan=\"4\"># of NVEF NVEF sense accuracy Using LS-NVWF a Using EWL b</td><td>Using Both c</td></tr><tr><td>4,028</td><td>74.8%(333/445)</td><td>87.6%(390/445)</td><td colspan=\"2\">89.2%(397/445) 93.7%(417/445)</td></tr></table>", | |
| "text": "Results of the WSD experiment for 445 Chinese un-segmented test sentences." | |
| } | |
| } | |
| } | |
| } |