Knowledge Graph and Corpus Driven Segmentation and Answer Inference for Telegraphic Entity-seeking Queries

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1104–1114, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Knowledge Graph and Corpus Driven Segmentation and Answer Inference for Telegraphic Entity-seeking Queries Mandar Joshi ∗ IBM Research mandarj90@in.ibm.com Uma Sawant IIT Bombay, Yahoo Labs uma@cse.iitb.ac.in Soumen Chakrabarti IIT Bombay soumen@cse.iitb.ac.in Abstract Much recent work focuses on formal in- terpretation of natural question utterances, with the goal of executing the resulting structured queries on knowledge graphs (KGs) such as Freebase. Here we address two limitations of this approach when ap- plied to open-domain, entity-oriented Web queries. First, Web queries are rarely well- formed questions. They are “telegraphic”, with missing verbs, prepositions, clauses, case and phrase clues. Second, the KG is always incomplete, unable to directly an- swer many queries. We propose a novel technique to segment a telegraphic query and assign a coarse-grained purpose to each segment: a base entity e1, a rela- tion type r, a target entity type t2, and contextual words s. The query seeks en- tity e2 ∈ t2 where r(e1, e2) holds, fur- ther evidenced by schema-agnostic words s. Query segmentation is integrated with the KG and an unstructured corpus where mentions of entities have been linked to the KG. We do not trust the best or any specific query segmentation. Instead, evi- dence in favor of candidate e2s are aggre- gated across several segmentations. Ex- tensive experiments on the ClueWeb cor- pus and parts of Freebase as our KG, us- ing over a thousand telegraphic queries adapted from TREC, INEX, and Web- Questions, show the efficacy of our ap- proach. For one benchmark, MAP im- proves from 0.2–0.29 (competitive base- lines) to 0.42 (our system). NDCG@10 improves from 0.29–0.36 to 0.54. ∗Work done as Masters student at IIT Bombay 1 Introduction A majority of Web queries mention an entity or type (Lin et al., 2012), as users increasingly ex- plore the Web of objects using Web search. To better support entity-oriented queries, commercial Web search engines are rapidly building up large catalogs of types, entities and relations, popu- larly called a “knowledge graph” (KG) (Gallagher, 2012). Despite these advances, robust, Web-scale, open-domain, entity-oriented search faces many challenges. Here, we focus on two. 1.1 “Telegraphic” queries First, the surface utterances of entity-oriented Web queries are dramatically different from TREC- or Watson-style factoid question answering (QA), where questions are grammatically well-formed. Web queries are usually “telegraphic”: they are short, rarely use function words, punctuations or clausal structure, and use relatively flexible word orders. E.g., the natural utterance “on the bank of which river is the Hermitage Museum lo- cated” may be translated to the telegraphic Web query hermitage museum river bank. Even on well-formed question utterances, 50% of in- terpretation failures are contributed by parsing or structural matching failures (Kwiatkowski et al., 2013). Telegraphic utterances will generally be even more challenging. Consequently, whereas TREC-QA/NLP-style research has focused on parsing and precise in- terpretation of a well-formed query sentence to a strongly …

Content truncated to 500 words