Learning Discourse-Level Structures for Question Answering

  • Boris Galitsky


Traditional parse trees are combined together and enriched with anaphora and rhetoric information to form a unified representation for a paragraph of text. We refer to these representations as parse thickets. They are introduced to support answering complex questions, which include multiple sentences, to tackle as many constraints expressed in this question as possible. The question answering system is designed so that an initial set of answers, which is obtained by a TF*IDF or other keyword search model, is re-ranked. Passage re-ranking is performed using matching of the parse thickets of answers with the parse thicket of the question. To do that, a graph representation and matching technique for parse structures for paragraphs of text have been developed. We define the operation of generalization of two parse thickets as a measure of semantic similarity between paragraphs of text to be the maximal common sub-thicket of these parse thickets.

Passage re-ranking improvement via parse thickets is evaluated in a variety of chatbot question-answering domains with long questions. Using parse thickets improves search accuracy compared with the bag-of words, the pairwise matching of parse trees for sentences, and the tree kernel approaches. As a baseline, we use a web search engine API, which provides much more accurate search results than the majority of search benchmarks, such as TREC. A comparative analysis of the impact of various sources of discourse information on the search accuracy is conducted. An open source plug-in for SOLR is developed so that the proposed technology can be easily integrated with industrial search engines.


  1. Aronovich L, Spiegler I (2007) CM-tree: a dynamic clustered index for similarity search in metric databases. Data Knowl Eng 63(3):919–946CrossRefGoogle Scholar
  2. Bar-Haim R, Dagan I, Greental I, Shnarch E (2007) Semantic inference at the lexical-syntactic level. In: AAAI’07 Proceedings of the 22nd national conference on artificial intelligence, Vancouver, BC, Canada, pp 871–876Google Scholar
  3. Barrena M, Jurado E, Márquez-Neila P, Pachón C (2010) A flexible framework to ease nearest neighbor search in multidimensional data spaces. Data Knowl Eng 69(1):116–136CrossRefGoogle Scholar
  4. Bertossi L, Chomicki J (2004) Query answering in inconsistent databases. In: Logics for emerging applications of databases. Springer, Berlin/Heidelberg, pp 43–83CrossRefGoogle Scholar
  5. Boulos J, Dalvi N, Mandhani B, Mathur S, Re C, Suciu D (2005) MYSTIQ: a system for finding more answers by using probabilities. SIGMOD, June 14–16, 2005, Baltimore, MD, USAGoogle Scholar
  6. Bouquet P, Kuper G, Scoz M, Zanobini S (2004) Asking and answering semantic queries. Meaning Coordination and Negotiation (MCN-04) at ISWC-2004, Hiroshima, JapanGoogle Scholar
  7. Bron C, Kerbosch J (1973) Algorithm 457: finding all cliques of an undirected graph. Commun ACM (ACM) 16(9):575–577CrossRefGoogle Scholar
  8. Calvanese D, De Giacomo G, Lembo D, Lenzerini M, Rosati R (2007) Tractable reasoning and efficient query answering in description logics: the DL-lite family. J Autom Reason 39:385–429MathSciNetCrossRefGoogle Scholar
  9. Chali Y, Joty SR, Hasan SA (2009) Complex question answering: unsupervised learning approaches and experiments. J Artif Intell Res 35:1–47MathSciNetCrossRefGoogle Scholar
  10. Clark S, Curran JR (2004) Parsing the WSJ using CCG and log-linear models. In: 42nd ACL, Barcelona, SpainGoogle Scholar
  11. Collins M, Duffy N (2002) Convolution kernels for natural language. In: Proceedings of NIPS, pp 625–632Google Scholar
  12. Costa d, André L, Carvalho ES d M, da Silva AS, Berlt K, Bezerra A (2007) A cost-effective method for detecting web site replicas on search engine databases. Data Knowl Eng 62(3):421–437. CrossRefGoogle Scholar
  13. Curran JR, Clark S, Bos J (2007) Linguistically motivated large-scale NLP with C&C and boxer. In: Proceedings of the ACL 2007 demonstrations session (ACL-07 demo), pp 33–36Google Scholar
  14. Düsterhöft A, Thalheim B (2004) Linguistic based search facilities in snowflake-like database schemes. Data Knowl Eng 48(2):177–198CrossRefGoogle Scholar
  15. Galitsky B (2003) Natural language question answering system: technique of semantic headers. Advanced Knowledge International, MagillGoogle Scholar
  16. Galitsky B (2012) Machine learning of syntactic parse trees for search and classification of text. Eng Appl AI 26(3):1072–1091Google Scholar
  17. Galitsky B (2017a) Improving relevance in a content pipeline via syntactic generalization. Eng Appl Artif Intell 58:1–26CrossRefGoogle Scholar
  18. Galitsky B (2017b) Matching parse thickets for open domain question answering. Data Knowl Eng 107:24–50CrossRefGoogle Scholar
  19. Galitsky B, Kuznetsov S (2008) Learning communicative actions of conflicting human agents. J Exp Theor Artif Intell 20(4):277–317CrossRefGoogle Scholar
  20. Galitsky B, Lebedeva N (2015) Recognizing documents versus meta-documents by tree kernel learning. In: FLAIRS conference, pp 540–545Google Scholar
  21. Galitsky B, González MP, Chesñevar CI (2009) A novel approach for classifying customer complaints through graphs similarities in argumentative dialogue. Decis Support Syst 46(3):717–729CrossRefGoogle Scholar
  22. Galitsky B, Dobrocsi G, de la Rosa JL, Kuznetsov SO (2010) From generalization of syntactic parse trees to conceptual graphs. In: Croitoru M, Ferré S, Lukose D (eds) Conceptual structures: from information to intelligence, 18th international conference on conceptual structures, ICCS 2010. Lecture notes in artificial intelligence, vol 6208, pp 185–190Google Scholar
  23. Galitsky B, Dobrocsi G, de la Rosa JL, Sergei O (2011) Kuznetsov: using generalization of syntactic parse trees for taxonomy capture on the web. 19th international conference on conceptual structures, ICCS 2011, pp 104–117Google Scholar
  24. Galitsky B, de la Rosa JL, Dobrocsi G (2012) Inferring the semantic properties of sentences by mining syntactic parse trees. Data Knowl Eng 81–82:21–45CrossRefGoogle Scholar
  25. Galitsky B, Usikov D, Sergei O (2013) Kuznetsov: parse thicket representations for answering multi-sentence questions. In: 20th international conference on conceptual structures, ICCS 2013, Hissar, Bulgaria, pp 285–293Google Scholar
  26. Galitsky B, Ilvovsky D, Kuznetsov S (2015) Rhetoric map of an answer to compound queries. ACL, Beijing, China, vol 2, pp 681–686Google Scholar
  27. Harabagiu S, Lacatusu F, Hickl A (2006) Answering complex questions with random walk models. In: Proceedings of the 29th annual international ACM SIGIR conference on Research and development in information retrieval (SIGIR ’06). ACM, New York, NY, USA, pp 220–227Google Scholar
  28. Haussler D (1999) Convolution kernels on discrete structures. Technical report ucs-crl-99-10, University of California Santa CruzGoogle Scholar
  29. Hong JL, Siew E-G, Egerton S (2010) Information extraction for search engines using fast heuristic techniques. Data Knowl Eng 69(2):169–196CrossRefGoogle Scholar
  30. Horrocks I, Tessaris S (2002) Querying the semantic web: a formal approach. The semantic web—ISWC 2002. Springer, Berlin/Heidelberg, pp 177–191Google Scholar
  31. Ilvovsky D (2014) Going beyond sentences when applying tree kernels. ACL student workshop, pp 56–63Google Scholar
  32. Jansen P, Surdeanu M, Clark P (2014) Discourse complements lexical semantics for non-factoid answer reranking. In: Proceedings of the 52nd annual meeting of the Association for Computational Linguistics (ACL), 2014, Baltimore, MD, USAGoogle Scholar
  33. Joty SR, Carenini G, Ng RT, Mehdad Y (2013) Combining intra-and multi- sentential rhetorical parsing for document-level discourse analysis. In: ACL, vol 1, pp 486–496Google Scholar
  34. Joty S, Moschitti A (2014) Discriminative reranking of discourse parses using tree kernels. In: Proceedings of the conference on empirical methods in natural language processing (EMNLP 2014), Doha, QatarGoogle Scholar
  35. Kamp HA (1981) Theory of truth and semantic representation. In: Groenendijk JAG, Janssen TMV, Stokhof MBJ (eds) Formal methods in the study of language. Mathematisch Centrum, AmsterdamGoogle Scholar
  36. Kim J-J, Pezik P, Rebholz-Schuhmann D (2008) MedEvi: retrieving textual evidence of relations between biomedical concepts from Medline. Bioinformatics 24(11):1410–1412CrossRefGoogle Scholar
  37. Kohavi R (1995) A study of cross-validation and bootstrap for accuracy estimation and model selection. In: International joint conference on artificial intelligence IJCAI 1995, Morgan Kaufmann Publishers Inc., San FranciscoGoogle Scholar
  38. Lehrer A (1974) Semantic fields and lexical structure. Benjamins, AmsterdamGoogle Scholar
  39. Lei Y, Uren V, Motta E (2006) Semsearch: a search engine for the semantic web. In: Managing knowledge in a world of networks. Lecture notes in computer science, vol 4248, pp 238–245Google Scholar
  40. Li X, Roth D (2002) Learning question classifiers. In: Proceedings of the 19th international conference on computational linguistics - volume 1 (COLING ’02), vol 1. Association for Computational Linguistics, Stroudsburg, pp 1–7Google Scholar
  41. Mann WC, Taboada M (2015.) Last downloaded 13 June 2015
  42. Mann WC, Thompson SA (1988) Rhetoric al structure theory: toward a functional theory of text organization. Text 8(3):243–281CrossRefGoogle Scholar
  43. Mann WC, Matthiessen CMIM, Thompson SA (1992) Rhetorical structure theory and text analysis. In: Mann WC, Thompson SA (eds) Discourse description: diverse linguistic analyses of a fund-raising text. John Benjamins, Amsterdam, pp 39–78CrossRefGoogle Scholar
  44. Mecca G, Raunich S, Pappalardo A (2007) A new algorithm for clustering search results. Data Knowl Eng 62(3):504–522CrossRefGoogle Scholar
  45. Moschitti A (2006) Efficient convolution kernels for dependency and constituent syntactic trees. In: Proceedings of the 17th European conference on machine learning, Berlin, GermanyGoogle Scholar
  46. Moschitti A, Quarteroni S (2011) Linguistic kernels for answer re-ranking in question answering systems. Inf Process Manag 47(6):825–842CrossRefGoogle Scholar
  47. Natsev A, Milind R (2005) Naphade Jelena Tesic. Learning the semantics of multimedia queries and concepts from a small number of examples. MM’05, November 6–11, 2005, SingaporeGoogle Scholar
  48. Palmer M (2009) Semlink: linking PropBank, VerbNet and FrameNet. In: Proceedings of the generative lexicon conference. September 2009, Pisa, Italy, GenLex-09Google Scholar
  49. Punyakanok V, Roth D, Yih W (2005) The necessity of syntactic parsing for semantic role labeling. IJCAI-05, Edinburgh, Scotland, UK, pp 1117–1123Google Scholar
  50. Searle J (1969) Speech acts: an essay in the philosophy of language. Cambridge University, CambridgeCrossRefGoogle Scholar
  51. Seo J, Simmons RF (1989) Syntactic graphs: a representation for the union of all ambiguous parse trees. Comput Linguist 15:15Google Scholar
  52. Severyn A, Moschitti A (2012) Fast support vector machines for convolution tree kernels. Data Min Knowl Disc 25:325–357MathSciNetCrossRefGoogle Scholar
  53. Sidorov G, Velasquez F, Stamatatos E, Gelbukh A, Chanona-Hernández L (2012) Syntactic dependency-based N-grams as classification features. LNAI 7630, pp 1–11Google Scholar
  54. Sidorov G, Velasquez F, Stamatatos E, Gelbukh A, Chanona-Hernández L (2013) Syntactic N-grams as machine learning features for natural language processing. Expert Syst Appl 41(3):853–860CrossRefGoogle Scholar
  55. Steedman M (2000) The syntactic process. The MIT Press, Cambridge, MAzbMATHGoogle Scholar
  56. Sun J, Zhang M, Tan C (2011) Tree sequence kernel for natural language. AAAI-25Google Scholar
  57. Tran T, Cimiano P, Rudolph S, Studer R (2007) Ontology-based interpretation of keywords for semantic search in “The semantic web”. Lecture notes in computer science, vol 4825, pp 523–536Google Scholar
  58. van Eijck J, Kamp H (1997) Representing discourse in context. Handbook of logic and language. Elsevier, Amsterdam, pp 179–237CrossRefGoogle Scholar
  59. Varlamis I, Stamou S (2009) Semantically driven snippet selection for supporting focused web searches. Data Knowl Eng 68(2):261–277. CrossRefGoogle Scholar
  60. Vo NPA, Popescu O (2016) A multi-layer system for semantic textual similarity. In: 8th international conference on knowledge discovery and information retrieval, vol 1, pp 56–67Google Scholar
  61. Vo NPA, Popescu O (2019) Multi-layer and co-learning systems for semantic textual similarity, semantic relatedness and recognizing textual entailment. In: 8th international joint conference, IC3K 2016, Porto, Portugal, November 9–11, 2016, Revised selected papers, pp 54–77Google Scholar
  62. Wu J, Xuan Z, Pan D (2011) Enhancing text representation for classification tasks with semantic graph structures. Int J Innov Comput Inf Control (ICIC) 7(5(B)):2689–2698Google Scholar
  63. Zhang M, Che W, Zhou G, Aw A, Tan C, Liu T, Li S (2008) Semantic role labeling using a grammar-driven convolution tree kernel. IEEE Trans Audio Speech Lang Process 16(7):1315–1329CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Boris Galitsky
    • 1
  1. 1.Oracle (United States)San JoseUSA

Personalised recommendations