Abstract: Script is the structured knowledge representation of prototypical real-life event sequences. Learning the commonsense knowledge inside the script can be helpful for machines in understanding natural language and drawing commonsensible inferences. script learning is an interesting and promising research direction, in which a trained script learning system can process narrative texts to capture script knowledge and draw inferences. However, there are currently no survey articles on script learning, so we are providing this comprehensive survey to deeply investigate the standard framework and the major research topics on script learning. This research field contains three main topics: event representations, script learning models, and evaluation approaches. For each topic, we systematically summarize and categorize the existing script learning systems, and carefully analyze and compare the advantages and disadvantages of the representative systems. We also discuss the current state of the research and possible future directions.

脚本学习综述

[1]Arrieta AB, Díaz-Rodríguez N, Del Ser J, et al., 2020. Explainable artificial intelligence (XAI): concepts, taxonomies, opportunities and challenges toward responsible AI. Inform Fus, 58:82-115.

[2]Balasubramanian N, Soderland S, Mausam, et al., 2013. Generating coherent event schemas at scale. Proc Conf on Empirical Methods in Natural Language Processing, p.1721-1731.

[3]Baroni M, Zamparelli R, 2010. Nouns are vectors, adjectives are matrices: representing adjective-noun constructions in semantic space. Proc Conf on Empirical Methods in Natural Language Processing, p.1183-1193.

[4]Bengio Y, Ducharme R, Vincent P, et al., 2003. A neural probabilistic language model. J Mach Learn Res, 3:1137-1155.

[5]Bordes A, Usunier N, Garcia-Durán A, et al., 2013. Translating embeddings for modeling multi-relational data. Proc 26^th Int Conf on Neural Information Processing Systems, p.2787-2795.

[6]Bower GH, Black JB, Turner TJ, 1979. Scripts in memory for text. Cogn Psychol, 11(2):177-220.

[7]Chambers N, 2017. Behind the scenes of an evolving event cloze test. Proc 2^nd Workshop on Linking Models of Lexical, Sentential and Discourse-Level Semantics, p.41-45.

[8]Chambers N, Jurafsky D, 2008. Unsupervised learning of narrative event chains. Proc 46^th Annual Meeting of the Association for Computational Linguistics, p.789-797.

[9]Chambers N, Jurafsky D, 2009. Unsupervised learning of narrative schemas and their participants. Proc Joint Conf of the 47^th Annual Meeting of the ACL and the 4^th Int Joint Conf on Natural Language Processing of the AFNLP, p.602-610.

[10]Chung J, Gulcehre C, Cho K, et al., 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. https://arxiv.org/abs/1412.3555

[11]Church KW, Hanks P, 1990. Word association norms, mutual information, and lexicography. Comput Ling, 16(1):22-29.

[12]Cullingford RE, 1978. Script Application: Computer Understanding of Newspaper Stories. PhD Thesis, Yale University, New Haven, CT, USA.

[13]DeJong GF, 1979. Skimming Stories in Real Time: an Experiment in Integrated Understanding. PhD Thesis, Yale University, New Haven, CT, USA.

[14]Devlin J, Chang MW, Lee K, et al., 2019. BERT: pre-training of deep bidirectional transformers for language understanding. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.4171-4186.

[15]Ding X, Li ZY, Liu T, et al., 2019a. ELG: an event logic graph. https://arxiv.org/abs/1907.08015

[16]Ding X, Liao K, Liu T, et al., 2019b. Event representation learning enhanced with external commonsense knowledge. Proc Conf on Empirical Methods in Natural Language Processing and the 9^th Int Joint Conf on Natural Language Processing, p.4896-4905.

[17]Erk K, Padó S, 2008. A structured vector space model for word meaning in context. Proc Conf on Empirical Methods in Natural Language Processing, p.897-906.

[18]Fillmore CJ, 1976. Frame semantics and the nature of language. Ann N Y Acad Sci, 280(1):20-32.

[19]Glavav s G, v Snajder J, 2015. Construction and evaluation of event graphs. Nat Lang Eng, 21(4):607-652.

[20]Gordon AS, 2001. Browsing image collections with representations of common-sense activities. J Am Soc Inform Sci Technol, 52(11):925-929.

[21]Granroth-Wilding M, Clark S, 2016. What happens next? Event prediction using a compositional neural network model. Proc 30^th AAAI Conf on Artificial Intelligence, p.2727-2733.

[22]Gupta R, Kochenderfer MJ, 2004. Common sense data acquisition for indoor mobile robots. Proc 19^th National Conf on Artifical Intelligence, p.605-610.

[23]Harris ZS, 1954. Distributional structure. Word, 10(2-3):146-162.

[24]Hochreiter S, Schmidhuber J, 1997. Long short-term memory. Neur Comput, 9(8):1735-1780.

[25]Hu LM, Li JZ, Nie LQ, et al., 2017. What happens next? Future subevent prediction using contextual hierarchical LSTM. Proc 31^st AAAI Conf on Artificial Intelligence, p.3450-3456.

[26]Jans B, Bethard S, Vulic, et al., 2012. Skip N-grams and ranking functions for predicting script events. Proc 13^th Conf of the European Chapter of the Association for Computational Linguistics, p.336-344.

[27]Jones MP, Martin JH, 1997. Contextual spelling correction using latent semantic analysis. Proc 5^th Conf on Applied Natural Language Processing, p.166-173.

[28]Kaelbling LP, Littman ML, Moore AW, 1996. Reinforcement learning: a survey. J Artif Intell Res, 4:237-285.

[29]Khan A, Salim N, Kumar YJ, 2015. A framework for multi-document abstractive summarization based on semantic role labelling. Appl Soft Comput, 30:737-747.

[30]Kiros R, Zhu YK, Salakhutdinov R, et al., 2015. Skip-thought vectors. Proc 28^th Int Conf on Neural Information Processing Systems, p.3294-3302.

[31]Koh PW, Liang P, 2017. Understanding black-box predictions via influence functions. Proc 34^th Int Conf on Machine Learning, p.1885-1894.

[32]Laender AHF, Ribeiro-Neto BA, Da Silva AS, et al., 2002. A brief survey of web data extraction tools. ACM SIGMOD Rec, 31(2):84-93.

[33]Lee G, Flowers M, Dyer MG, 1992. Learning distributed representations of conceptual knowledge and their application to script-based story processing. In: Sharkey N (Ed.), Connectionist Natural Language Processing. Springer, Dordrecht, p.215-247.

[34]Lee IT, Goldwasser D, 2018. FEEL: featured event embedding learning. Proc 32^nd AAAI Conf on Artificial Intelligence.

[35]Lee IT, Goldwasser D, 2019. Multi-relational script learning for discourse relations. Proc 57^th Annual Meeting of the Association for Computational Linguistics, p.4214-4226.

[36]Li JW, Monroe W, Ritter A, et al., 2016. Deep reinforcement learning for dialogue generation. Proc Conf on Empirical Methods in Natural Language Processing, p.1192-1202.

[37]Li Q, Li ZW, Wei JM, et al., 2018. A multi-attention based neural network with external knowledge for story ending predicting task. Proc 27^th Int Conf on Computational Linguistics, p.1754-1762.

[38]Li ZY, Ding X, Liu T, 2018. Constructing narrative event evolutionary graph for script event prediction. Proc 27^th Int Joint Conf on Artificial Intelligence, p.4201-4207.

[39]Li ZY, Ding X, Liu T, 2019. Story ending prediction by transferable BERT. Proc 28^th Int Joint Conf on Artificial Intelligence, p.1800-1806.

[40]Lin YK, Liu ZY, Sun MS, et al., 2015. Learning entity and relation embeddings for knowledge graph completion. Pro 29^th AAAI Conf on Artificial Intelligence.

[41]Lin ZH, Feng MW, Dos Santos CN, et al., 2017. A structured self-attentive sentence embedding. Proc 5^th Int Conf on Learning Representations.

[42]Luong T, Pham H, Manning CD, 2015. Effective approaches to attention-based neural machine translation. Proc Conf on Empirical Methods in Natural Language Processing, p.1412-1421.

[43]Lv SW, Qian WH, Huang LT, et al., 2019. SAM-Net: integrating event-level and chain-level attentions to predict what happens next. Proc AAAI Conf on Artificial Intelligence, p.6802-6809.

[44]Mausam, Schmitz M, Bart R, et al., 2012. Open language learning for information extraction. Proc Joint Conf on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, p.523-534.

[45]McCann B, Bradbury J, Xiong CM, et al., 2017. Learned in translation: contextualized word vectors. Proc 31^st Int Conf on Neural Information Processing Systems, p.6297-6308.

[46]Miikkulainen R, 1992. Discern: a distributed neural network model of script processing and memory. University Twente, Connectionism and Natural Language Processing, p.115-124.

[47]Miikkulainen R, 1993. Subsymbolic Natural Language Processing: an Integrated Model of Scripts, Lexicon, and Memory. MIT Press, Cambridge, USA.

[48]Mikolov T, Chen K, Corrado G, et al., 2013. Efficient estimation of word representations in vector space. https://arxiv.org/abs/1301.3781

[49]Miller GA, 1995. WordNet: a lexical database for English. Commun ACM, 38(11):39-41.

[50]Minsky M, 1975. A framework for representing knowledge. In: Winston PH (Ed.), The Psychology of Computer Vision. McGraw-Hill Book, New York, USA.

[51]Mnih A, Hinton G, 2007. Three new graphical models for statistical language modelling. Proc 24^th Int Conf on Machine Learning, p.641-648.

[52]Modi A, 2016. Event embeddings for semantic script modeling. Proc 20^th SIGNLL Conf on Computational Natural Language Learning, p.75-83.

[53]Modi A, Titov I, 2014a. Inducing neural models of script knowledge. Proc 18^th Conf on Computational Natural Language Learning, p.49-57.

[54]Modi A, Titov I, 2014b. Learning semantic script knowledge with event embeddings. Proc 2^nd Int Conf on Learning Representations.

[55]Modi A, Anikina T, Ostermann S, et al., 2016. InScript: narrative texts annotated with script information. Proc 10^th Int Conf on Language Resources and Evaluation.

[56]Modi A, Titov I, Demberg V, et al., 2017. Modeling semantic expectation: using script knowledge for referent prediction. Trans Assoc Comput Ling, 5(2):31-44.

[57]Mostafazadeh N, Chambers N, He XD, et al., 2016. A corpus and cloze evaluation for deeper understanding of commonsense stories. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.839-849.

[58]Mueller ET, 1998. Natural Language Processing with ThoughtTreasure. Signiform, New York, USA.

[59]Navigli R, 2009. Word sense disambiguation: a survey. ACM Comput Surv, 41(2):10.

[60]Orr JW, Tadepalli P, Doppa JR, et al., 2014. Learning scripts as hidden Markov models. Proc 28^th AAAI Conf on Artificial Intelligence, p.1565-1571.

[61]Osman AH, Salim N, Binwahlan MS, et al., 2012. Plagiarism detection scheme based on semantic role labeling. Proc Int Conf on Information Retrieval & Knowledge Management, p.30-33.

[62]Pei KX, Cao YZ, Yang JF, et al., 2017. DeepXplore: automated whitebox testing of deep learning systems. Proc 26^th Symp on Operating Systems Principles, p.1-18.

[63]Pennington J, Socher R, Manning C, 2014. GloVe: global vectors for word representation. Proc Conf on Empirical Methods in Natural Language Processing, p.1532-1543.

[64]Perozzi B, Al-Rfou R, Skiena S, 2014. DeepWalk: online learning of social representations. Proc 20^th ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining, p.701-710.

[65]Peters M, Neumann M, Iyyer M, et al., 2018. Deep contextualized word representations. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.2227-2237.

[66]Pichotta K, Mooney R, 2014. Statistical script learning with multi-argument events. Proc 14^th Conf of the European Chapter of the Association for Computational Linguistics, p.220-229.

[67]Pichotta K, Mooney RJ, 2016a. Learning statistical scripts with LSTM recurrent neural networks. Proc 30^th AAAI Conf on Artificial Intelligence, p.2800-2806.

[68]Pichotta K, Mooney RJ, 2016b. Using sentence-level LSTM language models for script inference. Proc 54^th Annual Meeting of the Association for Computational Linguistics, p.279-289.

[69]Prasad R, Dinesh N, Lee A, et al., 2008. The Penn discourse Treebank 2.0. Proc Int 6^th Conf on Language Resources and Evaluation, p.2961-2968.

[70]Qiu XP, Sun TX, Xu YG, et al., 2020. Pre-trained models for natural language processing: a survey. https://arxiv.org/abs/2003.08271

[71]Radford A, Narasimhan K, Salimans T, et al., 2019. Improving language understanding by generative pre-training. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.4171-4186.

[72]Radinsky K, Agichtein E, Gabrilovich E, et al., 2011. A word at a time: computing word relatedness using temporal semantic analysis. Proc 20^th Int Conf on World Wide Web, p.337-346.

[73]Rashkin H, Sap M, Allaway E, et al., 2018. Event2Mind: commonsense inference on events, intents, and reactions. Proc 56^th Annual Meeting of the Association for Computational Linguistics, p.463-473.

[74]Regneri M, Koller A, Pinkal M, 2010. Learning script knowledge with web experiments. Proc 48^th Annual Meeting of the Association for Computational Linguistics, p.979-988.

[75]Rudinger R, Rastogi P, Ferraro F, et al., 2015. Script induction as language modeling. Proc Conf on Empirical Methods in Natural Language Processing, p.1681-1686.

[76]Rumelhart DE, 1980. Schemata: the building blocks of cognition. In: Spiro RJ (Ed.), Theoretical Issues in Reading Comprehension. Erlbaum, Hillsdale, p.33-58.

[77]Sap M, Le Bras R, Allaway E, et al., 2019. ATOMIC: an atlas of machine commonsense for if-then reasoning. Proc AAAI Conf on Artificial Intelligence, p.3027-3035.

[78]Schank RC, 1983. Dynamic Memory: a Theory of Reminding and Learning in Computers and People. Cambridge University Press, New York, USA.

[79]Schank RC, 1990. Tell Me a Story: a New Look at Real and Artificial Memory. Charles Scribner, New York, USA.

[80]Schank RC, Abelson RP, 1977. Scripts, Plans, Goals and Understanding: an Inquiry into Human Knowledge Structures. L. Erlbaum, Hillsdale, USA.

[81]Schuler KK, 2005. VerbNet: a Broad-Coverage, Comprehensive Verb Lexicon. PhD Thesis, University of Pennsylvania, Pennsylvania, USA.

[82]Shen D, Lapata M, 2007. Using semantic roles to improve question answering. Proc Joint Conf on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, p.12-21.

[83]Socher R, Huval B, Manning CD, et al., 2012. Semantic compositionality through recursive matrix-vector spaces. Proc Joint Conf on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, p.1201-1211.

[84]Sutton RS, Barto AG, 2018. Reinforcement Learning: an Introduction (2^nd Ed.). MIT Press, Cambridge, USA.

[85]Taylor WL, 1953. “Cloze procedure”: a new tool for measuring readability. J Mass Commun Q, 30(4):415-433.

[86]Terry WS, 2006. Learning and Memory: Basic Principles, Processes, and Procedures. Allyn and Bacon, Boston, USA.

[87]Tulving E, 1983. Elements of Episodic Memory. Oxford University Press, New York, USA.

[88]Wang Z, Zhang JW, Feng JL, et al., 2014. Knowledge graph embedding by translating on hyperplanes. Proc 28^th AAAI Conf on Artificial Intelligence, p.1112-1119.

[89]Wang ZQ, Zhang Y, Chang CY, 2017. Integrating order information and event relation for script event prediction. Proc Conf on Empirical Methods in Natural Language Processing, p.57-67.

[90]Weber N, Balasubramanian N, Chambers N, 2018. Event representations with tensor-based compositions. Proc 32^nd AAAI Conf on Artificial Intelligence, p.4946-4953.

[91]Weston J, Chopra S, Bordes A, 2015. Memory networks. https://arxiv.org/abs/1410.3916

[92]Zhao SD, Wang Q, Massung S, et al., 2017. Constructing and embedding abstract event causality networks from text snippets. Proc 10^th ACM Int Conf on Web Search and Data Mining, p.335-344.

[93]Zheng JM, Cai F, Chen HH, 2020. Incorporating scenario knowledge into a unified fine-tuning architecture for event representation. Proc 43^rd Int ACM SIGIR Conf on Research and Development in Information Retrieval, p.249-258.

[94]Zhou MT, Huang ML, Zhu XY, 2019. Story ending selection by finding hints from pairwise candidate endings. IEEE/ACM Trans Audio Speech Lang Process, 27(4):719-729.