| { | |
| "paper_id": "2020", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T01:05:59.915864Z" | |
| }, | |
| "title": "The challenge of the TV game La Ghigliottina to NLP", | |
| "authors": [ | |
| { | |
| "first": "Federico", | |
| "middle": [], | |
| "last": "Sangati", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "UNIOR NLP Research Group", | |
| "institution": "Orientale University of Naples", | |
| "location": {} | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Antonio", | |
| "middle": [], | |
| "last": "Pascucci", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "UNIOR NLP Research Group", | |
| "institution": "Orientale University of Naples", | |
| "location": {} | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Johanna", | |
| "middle": [], | |
| "last": "Monti", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "UNIOR NLP Research Group", | |
| "institution": "Orientale University of Naples", | |
| "location": {} | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "In this paper, we describe a Telegram bot, Mago della Ghigliottina (Ghigliottina Wizard), able to solve La Ghigliottina game (The Guillotine), the final game of the Italian TV quiz show L'Eredit\u00e0. Our system relies on linguistic resources and artificial intelligence and achieves better results than human players (and competitors of L'Eredit\u00e0 too). In addition to solving a game, Mago della Ghigliottina can also generate new game instances and challenge the users to match the solution.", | |
| "pdf_parse": { | |
| "paper_id": "2020", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "In this paper, we describe a Telegram bot, Mago della Ghigliottina (Ghigliottina Wizard), able to solve La Ghigliottina game (The Guillotine), the final game of the Italian TV quiz show L'Eredit\u00e0. Our system relies on linguistic resources and artificial intelligence and achieves better results than human players (and competitors of L'Eredit\u00e0 too). In addition to solving a game, Mago della Ghigliottina can also generate new game instances and challenge the users to match the solution.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "In this paper, we present Mago della Ghigliottina, a Telegram bot able to solve La Ghigliottina (The Guillotine), the final game of the Italian TV quiz show L'Eredit\u00e0. Given a set of five words (clues), the competitor has to guess the sixth word (solution) that is linked with each of these five clues. For example, given the five clues: pie, bad, Adam, core, eye the solution is apple, because: apple-pie is a kind of pie; bad apple is a way to refer to a trouble maker; Adam's apple is the prominent part of men's throat; apple core is the center of the apple; apple of someone's eye is a way to refer to someone's beloved person. The game is one of the most popular games in Italy, viewed by nearly three million spectators (approximately 20% share). Audience at home also enjoys participating in the game and tries to match the solution. Some players compete in Ghigliottiniamo 1 , an app that enables people to challenge each other while the game is being broadcasted. Mago della Ghigliottina participated as UNIOR4NLP (Sangati et al., 2018) in the shared task NLP4FUN (Basile et al., 2018) , which was part of the EVALITA 2018 (Caselli et al., 2018) , a periodic evaluation campaign of Natural Language Processing (NLP) and speech tools for the Italian language. 2 The paper is organized as follows: in Section 2 we present La Ghigliottina TV game and its rules. In Section 3 we make a brief overview about NLP systems related to games. In Section 4 we describe our approach to solve the game: how we built our artificial player and how the system attempts at finding the correct solution. In Section 5 we describe Mago della Ghigliottina. Conclusions and Future Work are in Section 6.", | |
| "cite_spans": [ | |
| { | |
| "start": 1024, | |
| "end": 1046, | |
| "text": "(Sangati et al., 2018)", | |
| "ref_id": "BIBREF7" | |
| }, | |
| { | |
| "start": 1074, | |
| "end": 1095, | |
| "text": "(Basile et al., 2018)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 1133, | |
| "end": 1155, | |
| "text": "(Caselli et al., 2018)", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 1269, | |
| "end": 1270, | |
| "text": "2", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "L'Eredit\u00e0 represents one of the most popular TV games in Italy and in 2020 reached its 18 th season. The game involves seven competitors that challenge each other in elimination games. The player who reaches the final game (named La Ghigliottina) can win the jackpot. The game works as follows: given a set of five words (clues), the player has to guess a sixth word (the solution) that is linked with the five clues. The five clues are unrelated to each other, but each is in relation with the solution. In figure 1 we show an example of the game. Figure 1 : A screenshot of La Ghigliottina. In this case, the solution is cassa: i) cassa del cinema (cinema box office), ii) grancassa (bass drum), iii) cassa comune (petty cash), iv) battere cassa (beat the check), and v) coda alla cassa (checkout line) .", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 549, | |
| "end": 557, | |
| "text": "Figure 1", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "La Ghigliottina game", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "Games represent an interesting playground to conduct research in NLP and Artificial Intelligence (AI) (Yannakakis and Togelius, 2018) . There are a number of popular TV quiz based on language games, such as The Wheel of Fortune and Who Wants to be a Millionaire?. The current work focuses on La Ghigliottina, which is particularly interesting for AI and NLP because it is based on how words are connected to each other. OTTHO (Semeraro et al., 2009; Basile et al., 2014) represents the first artificial player of the game and exploits resources from the web such as Wikipedia to build i) a lexicon and a knowledge repository and ii) a knowledge base modeling represented by an association matrix which stores the degree of correlation between any two words in the lexicon. Word correlations are detected by connecting lemmas to the entries in the dictionary definition, pair of words occurring in proverbs, movie or song titles, and pair of similar words by exploiting Vector Space Models (Salton et al., 1975) .", | |
| "cite_spans": [ | |
| { | |
| "start": 102, | |
| "end": 133, | |
| "text": "(Yannakakis and Togelius, 2018)", | |
| "ref_id": "BIBREF9" | |
| }, | |
| { | |
| "start": 426, | |
| "end": 449, | |
| "text": "(Semeraro et al., 2009;", | |
| "ref_id": "BIBREF8" | |
| }, | |
| { | |
| "start": 450, | |
| "end": 470, | |
| "text": "Basile et al., 2014)", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 989, | |
| "end": 1010, | |
| "text": "(Salton et al., 1975)", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "NLP and games", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Our approach uses similar resources with respect to the OT-THO system. However, regarding word relations we don't focus on word similarity but on a restricted set of syntactic constructions (patterns).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Our approach to solve La Ghigliottina", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Building an automatic solver for La Ghigliottina game requires three preliminary steps: game analysis, definition of linguistic patterns, and extraction of linguistic resources.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Preliminary steps in building our artificial player", | |
| "sec_num": "4.1." | |
| }, | |
| { | |
| "text": "We analyzed a sample of 100 La Ghigliottina instances that we collected from past editions of the TV game. We found out that almost in all combination the clue is connected to the solution because they form a Multiword Expression (MWE). A MWE is a sequence of words that presents some characteristic behaviour (at the lexical, syntactic, semantic, pragmatic or statistical level) and whose interpretation crosses the boundaries between words (Sag et al., 2002) . MWEs have to be considered as lexical items which convey a single meaning different from the meanings of its constituent words, such as in the idiomatic expression kick the bucket where the simple addition of the meanings of kick and bucket does not convey the meaning of to die. Our system has been built on this key observation. After the official dataset was released, we found out that the great majority of game instances confirmed our initial hypothesis.", | |
| "cite_spans": [ | |
| { | |
| "start": 442, | |
| "end": 460, | |
| "text": "(Sag et al., 2002)", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Linguistic Patterns Clue words are typically nouns, verbs, or adjectives, while the solutions are typically nouns or adjectives (almost never verbs). We detected six possible clue-solution combination that generate MWEs:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A-B (Noun-Adjective, Adjective-Noun, Verb-Noun, Noun-Noun) permesso premio ('permit price' \u2192 good behaviour license);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A-determiner-B dare il permesso ('give the permit' \u2192 authorize);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A-conjunction-B stima e affetto (esteem and affection);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A-preposition-B colpo di coda ('flick of tail' \u2192 last ditch effort);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A-articulated preposition-B virt\u00f9 dei forti, part of the famous Italian proverb La calma\u00e8 la virt\u00f9 dei forti (patience is the virtue of the strong);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 A+B: compounds such as radio + attivit\u00e0 = radioattivit\u00e0 (radio + activity = radioactivity).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Game analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We collected the linguistic resources which we deemed necessary for the task. To this end we used the following freely available corpora:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 Pais\u00e0: 250 million tokens corpus automatically annotated (Lyding et al., 2014).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 itWaC: 1.5 billion words corpus automatically annotated (Baroni et al., 2009) \u2022 Wiki-IT-Titles: Wikipedia-IT titles downloaded via WikiExtractor (Attardi, 2012 ).", | |
| "cite_spans": [ | |
| { | |
| "start": 58, | |
| "end": 79, | |
| "text": "(Baroni et al., 2009)", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 147, | |
| "end": 161, | |
| "text": "(Attardi, 2012", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 Proverbs: 1,955 proverbs from (Wikiquote, 2016) and 371 from an online collection (Dige, 2016) .", | |
| "cite_spans": [ | |
| { | |
| "start": 32, | |
| "end": 49, | |
| "text": "(Wikiquote, 2016)", | |
| "ref_id": "BIBREF16" | |
| }, | |
| { | |
| "start": 84, | |
| "end": 96, | |
| "text": "(Dige, 2016)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In addition, we have constructed the following lexical resources:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 DeMauro-Ext: words extracted from \"Il Nuovo vocabolario di base della lingua italiana\"(De Mauro, 2016b), extended with morphological variations obtained by changing last vowel of the word and checking if the resulting word has frequency \u2265 1000 in Pais\u00e0.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 DeMauro-MWEs: MWEs extracted from the \"De Mauro online dictionary\" (De Mauro, 2016a) composed of 30,633 entries.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistic Resources", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In order to build our system, we started processing the selected corpora via standard tokenization (only single word tokens) and removal of punctuation marks and non-word patterns. Next, we constructed two lexical sets: C LEX to cover the clue words, and S LEX to cover the solution words. S LEX (composed of 7,942 nouns and adjectives in DeMauro-Ext) is smaller than C LEX (composed of 19,414 words from the full DeMauro-Ext and DeMauro-MWEs) because solution words are almost always nouns or adjectives as described in Section 4.1. Secondly, we built a co-occurrence matrix M c which stores the counts c i,j for every pair of words w i \u2208 S LEX and w j \u2208 C LEX such that w i co-occurs with w j in the resources according to patterns described in Section 4.1. Co-occurrence patterns were extracted from Pais\u00e0 and itWaC with weight w = 1, from DeMauro-MWE with w = 200, from Proverbs with w = 100, and from Wiki-IT-Titles with w = 50. The weight were chosen manually taking into account the likelihood that a pattern in a given corpus represented a valid MWE. Compound patterns (A+B) were extracted from C LEX : for every word w in C LEX if w = ab, a and b are both in C LEX , and a and b have at least 4 characters, the count for the pair (a, b) is incremented by 1 in the co-occurrence matrix. Thirdly, for every pair of words w i and w j in M c , we populate the association-score matrix M pmi via the Pointwise Mutual Information measure:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "M pmi (w i , w j ) = log p(w i , w j ) p(w i ) \u2022 p(w j )", | |
| "eq_num": "(1)" | |
| } | |
| ], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "where", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "p(w i ) = wj \u2208C LEX M c (w i , w j ) (2) p(w j ) = wi\u2208S LEX M c (w i , w j ) (3) p(w i , w j ) = M c (w i , w j ) x\u2208S LEX y\u2208C LEX M c (x, y)", | |
| "eq_num": "(4)" | |
| } | |
| ], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Finally, for a given game instance with the 5 clue words G = (w c1 , w c2 , w c3 , w c4 , w c5 ), we choose the solution word w s \u2208 S LEX such that:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "EQUATION", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [ | |
| { | |
| "start": 0, | |
| "end": 8, | |
| "text": "EQUATION", | |
| "ref_id": "EQREF", | |
| "raw_str": "w s = max ws\u2208S LEX wc\u2208G M pmi (w s , w c )", | |
| "eq_num": "(5)" | |
| } | |
| ], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "that is, we choose the word in S LEX which maximizes the score obtained by summing the pmi between each clue word and the candidate word. If two words are never seen co-occurring together in a pattern in the training corpora, we assign to them the lowest pmi value in M pmi .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "The system has been implemented in Python and the code is open source. After the matrix has been loaded into memory the response time on an average laptop is around 1-2 seconds.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "System description", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Mago della Ghigliottina offers two game modalities: the first one (solution mode) allows users to insert the five clues and to challenge the bot to match the solution, while the second one (generation mode) presents the user with five clue words and challenges to find the correct solution.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mago della Ghigliottina bot", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "In solution mode, it is possible to write the five clues or to send a picture with the five clues. As shown in Figure 2 , Mago della Ghigliottina uses OCR to recognize the words from the image. Mago della Ghigliottina returns its prediction with a degree of accuracy accompanied by an emoji face that reflects the specific degree of accuracy:", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 111, | |
| "end": 119, | |
| "text": "Figure 2", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Solution mode", | |
| "sec_num": "5.1." | |
| }, | |
| { | |
| "text": "\u2022 sono quasi certo (I'm fairly certain, as shown in the example below).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Solution mode", | |
| "sec_num": "5.1." | |
| }, | |
| { | |
| "text": "credo che (I believe that);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 sono abbastanza convinto (I'm quite sure);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 non sono sicuro, ma (I'm not sure, but);", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Until now (April 2020), 9, 333 Ghigliottina instances (4016 unique) have been submitted to the bot by a total of 740 Telegram users and 133 Twitter users.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Mago della Ghigliottina is automatically tested every day on the TV show instances and is able to guess the solution correctly about 2/3 of the time. It must be considered that every day several users submit the same instance of La Ghigliottina game, so this performance has been calculated discarding duplicates. The bot definitely outperforms humans in solving the game. In comparing the performance of our AI system with that of a top player, we analyzed the games played by Andrea Saccone, who has been the biggest champion of the Ghigliottina game so far: he was champion for 13 days (3-15 March 2018), and he managed to Mago della Ghigliottina accepts the picture containing the five clues and in 1-2 seconds returns its prediction with a degree of accuracy. In this case, the system correctly guesses the solution because of the following MWEs i) conoscere alla perfezione (to perfectly know), ii) grado di perfezione (degree of perfection), iii) modello di perfezione (model of perfection), iv) ideale di perfezione (ideal of perfection), and v) perfezione divina (divine perfection).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "find the correct solution three times. 3 In comparison Mago della Ghigliottina was able to win the same game instances 9 times. As mentioned in section 3, the other AI model that has been developed to solve the same task is OTTHO (Basile et al., 2014) , which makes use of Vector Space Models (among other things). We observe that they achieve a precision of around 30% which is well below ours (75%). 4 Users can send live to Mago della Ghigliottina the Ghigliottina instance. Alternatively, users can send new instances made up on the spot whenever they want, just thinking at five words related to the word they imagine as the solution.", | |
| "cite_spans": [ | |
| { | |
| "start": 39, | |
| "end": 40, | |
| "text": "3", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 230, | |
| "end": 251, | |
| "text": "(Basile et al., 2014)", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 402, | |
| "end": 403, | |
| "text": "4", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Obviously, in this case prediction performaces can drop because pairs clue-solution can never be accurate as well as those of the TV show.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u2022", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In generation mode (Figure 3) , the system can automatically create novel game instances. Using a reversed association-matrix, 5 it chooses a random word (the solu-tion) and presents a list of 5 clues with a high score. In addition, when the solution is given it provides a sentence for every clue-solution pair exemplifying the relations between the two words. In this case, the solution is scherzo (joke): i) nemmeno per scherzo (even as a joke), ii) brutto scherzo (bad joke), iii) scherzo da prete (sick joke), iv) scherzo di natura (trick of nature), and v) neppure per scherzo (even as a joke)", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 19, | |
| "end": 29, | |
| "text": "(Figure 3)", | |
| "ref_id": "FIGREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Generation mode", | |
| "sec_num": "5.2." | |
| }, | |
| { | |
| "text": "In this paper, we have described Mago della Ghigliottina, a Telegram bot that is able to solve La Ghigliottina game (The Guillotine), the final game of the Italian TV quiz show L'Eredit\u00e0. Mago della Ghigliottina relies on linguistic resources and is tested every day on the TV show instances: it is able to guess on average two out of three Ghigliottina instances. Users can play with Mago della Ghigliottina also independently for the TV game by providing their own clues and asking the bot to solve the game (solution mode) or by asking the bot five new clues (generation mode). We aim to collect continuously data in order to develop a corpus with new game instances and improve our system. Mago della Ghigliottina is also available on Amazon Alexa 6 (simply ask Mago della Ghigliottina to match the solution providing it with the five clues) and Twitter 7 (users just need to write the five clues and tag @UNIOR4NLP in their tweet and the solution appears in a comment). The methodology adopted for the bot can be successfully applied to all NLP tasks that aim at identifying cooccurrence and semantic relations between words. As future work we intend to add a validation step by the users when they play with the bot independently from the TV game. The validation is carried out by asking users to confirm the correctness of the solution provided by the bot to their Ghigliottina instances. In addition, we also foresee to ask users to provide all the correct co-occurrences between their five clues and the solution they expect.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Future Work", | |
| "sec_num": "6." | |
| }, | |
| { | |
| "text": "higher quality game instances. 6 https://www.amazon.it/ Federico-Sangati-Mago-della-Ghigliottina/ dp/B07VHKT43F 7 https://twitter.com/UNIOR4NLP", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and Future Work", | |
| "sec_num": "6." | |
| }, | |
| { | |
| "text": "https://appadvice.com/app/ ghigliottiniamo/1447355292 2 http://www.evalita.it", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The players who reach the \"Ghigliottina\" game (the champion) continue to participate in the subsequent episodes even if they do not guess the solution word.4 Unfortunately we were not able to make an exact comparison because the original data-set has not been made publicly available for copyright reasons.5 In generation mode, we make use of a smaller co-occurrence matrix extracted only from DeMauro-MWEs, in order to produce", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "This research has been partly supported by the PON Ricerca e Innovazione 2014/20 fund. Authorship contribution is as follows: Johanna Monti is author of Sections 1, 2 and 6, Federico Sangati is author of Sections 3, 4.2 and 5, while Section 4.1 is in common between Antonio Pascucci and Johanna Monti.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Acknowledgements", | |
| "sec_num": "7." | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Solving a complex language game by using knowledge-based word associations discovery", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Basile", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "De Gemmis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Lops", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Semeraro", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "IEEE Transactions on Computational Intelligence and AI in Games", | |
| "volume": "8", | |
| "issue": "1", | |
| "pages": "13--26", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Basile, P., de Gemmis, M., Lops, P., and Semeraro, G. (2014). Solving a complex language game by using knowledge-based word associations discovery. IEEE Transactions on Computational Intelligence and AI in Games, 8(1):13-26.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Overview of the evalita 2018 solving language games (nlp4fun) task. EVALITA Evaluation of NLP and Speech Tools for Italian", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Basile", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "De Gemmis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Siciliani", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Semeraro", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "", | |
| "volume": "12", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Basile, P., de Gemmis, M., Siciliani, L., and Semeraro, G. (2018). Overview of the evalita 2018 solving language games (nlp4fun) task. EVALITA Evaluation of NLP and Speech Tools for Italian, 12:75.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Sixth evaluation campaign of natural language processing and speech tools for italian: Final workshop (evalita 2018)", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Caselli", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Novielli", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Patti", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Rosso", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "EVALITA 2018. CEUR Workshop Proceedings", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Caselli, T., Novielli, N., Patti, V., and Rosso, P. (2018). Sixth evaluation campaign of natural language process- ing and speech tools for italian: Final workshop (evalita 2018). In EVALITA 2018. CEUR Workshop Proceedings (CEUR-WS. org).", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Il Nuovo De Mauro (Online)", | |
| "authors": [ | |
| { | |
| "first": "De", | |
| "middle": [], | |
| "last": "Mauro", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "De Mauro, T. (2016a). Il Nuovo De Mauro (Online). https://dizionario.internazionale.it. Last accessed on the 1st October 2018.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Il Nuovo vocabolario di base della lingua italiana (pdf version", | |
| "authors": [ | |
| { | |
| "first": "De", | |
| "middle": [], | |
| "last": "Mauro", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "De Mauro, T. (2016b). Il Nuovo vocabolario di base della lingua italiana (pdf version). https: //www.dropbox.com/s/mkcyo53m15ktbnp/ nuovovocabolariodibase.pdf. Last accessed on the 1st October 2018.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Multiword expressions: A pain in the neck for nlp", | |
| "authors": [ | |
| { | |
| "first": "I", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Sag", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Baldwin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Bond", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Copestake", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Flickinger", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "International conference on intelligent text processing and computational linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--15", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sag, I. A., Baldwin, T., Bond, F., Copestake, A., and Flickinger, D. (2002). Multiword expressions: A pain in the neck for nlp. In International conference on intelli- gent text processing and computational linguistics, pages 1-15. Springer.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "A vector space model for automatic indexing", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Salton", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Wong", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yang", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C.-S", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1975, | |
| "venue": "Communications of the ACM", | |
| "volume": "18", | |
| "issue": "11", | |
| "pages": "613--620", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Salton, G., Wong, A., and Yang, C.-S. (1975). A vector space model for automatic indexing. Communications of the ACM, 18(11):613-620.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Exploiting multiword expressions to solve \"la ghigliottina", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Sangati", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Pascucci", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Monti", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "Sixth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2018)", | |
| "volume": "2263", | |
| "issue": "", | |
| "pages": "258--263", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sangati, F., Pascucci, A., and Monti, J. (2018). Exploiting multiword expressions to solve \"la ghigliottina\". In Sixth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2018), volume 2263, pages 258-263. Accademia Univer- sity Press.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "On the tip of my thought: Playing the guillotine game", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Semeraro", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Lops", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Basile", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "De", | |
| "middle": [], | |
| "last": "Gemmis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Twenty-First International Joint Conference on Artificial Intelligence", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Semeraro, G., Lops, P., Basile, P., and De Gemmis, M. (2009). On the tip of my thought: Playing the guillotine game. In Twenty-First International Joint Conference on Artificial Intelligence.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Artificial intelligence and games", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [ | |
| "N" | |
| ], | |
| "last": "Yannakakis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Togelius", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2018, | |
| "venue": "", | |
| "volume": "2", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yannakakis, G. N. and Togelius, J. (2018). Artificial intel- ligence and games, volume 2. Springer.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "The wacky wide web: a collection of very large linguistically processed web-crawled corpora. Language resources and evaluation", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Baroni", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Bernardini", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Ferraresi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Zanchetta", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "", | |
| "volume": "43", | |
| "issue": "", | |
| "pages": "209--226", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Baroni, M., Bernardini, S., Ferraresi, A., and Zanchetta, E. (2009). The wacky wide web: a collection of very large linguistically processed web-crawled corpora. Language resources and evaluation, 43(3):209-226.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "The paisa'corpus of italian web texts", | |
| "authors": [ | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Lyding", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Stemle", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Borghetti", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Brunello", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Castagnoli", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Dell'orletta", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Dittmann", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Lenci", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Pirrelli", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "EACL (European chapter of the Association for Computational Linguistics)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "36--43", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lyding, V., Stemle, E., Borghetti, C., Brunello, M., Castag- noli, S., Dell'Orletta, F., Dittmann, H., Lenci, A., and Pirrelli, V. (2014). The paisa'corpus of italian web texts. In 9th Web as Corpus Workshop (WaC-9)@ EACL 2014, pages 36-43. EACL (European chapter of the Associa- tion for Computational Linguistics).", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Proverbi italiani", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Wikiquote", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2016, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wikiquote. (2016). Proverbi italiani. https://it. wikiquote.org/wiki/Proverbi_italiani. Downloaded on the 24th April 2018.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "type_str": "figure", | |
| "num": null, | |
| "text": "A screenshot from Telegram in solution mode:", | |
| "uris": null | |
| }, | |
| "FIGREF1": { | |
| "type_str": "figure", | |
| "num": null, | |
| "text": "A screenshot from Telegram in generation mode.", | |
| "uris": null | |
| } | |
| } | |
| } | |
| } |