Question Answering with Subgraph Embeddings

Bordes, Antoine; Chopra, Sumit; Weston, Jason

doi:10.3115/v1/d14-1067

Cited by 589 publications

(420 citation statements)

References 8 publications

Supporting

Mentioning

419

Contrasting

Unclassified

Order By: Relevance

“…We use the detected topic entity mentions to obtain candidate matching entities in the KB using Freebase Search API. We use top-Model F1 (Berant et al, 2013) 35.7 (Yao and Van Durme, 2014) 33.0 (Berant and Liang, 2014) 39.9 (Bao et al, 2014) 37.5 (Bordes et al, 2014) 39.2 (Yang et al, 2014) 41.3 (Dong et al, 2015b) 40.8 (Yao, 2015) 44.3 (Berant and Liang, 2015) 49.7 52.5 50.3 (Xu et al, 2016) 53 3 entities returned for the pruning step of Question Abstraction on the test examples. Answer Type Prediction.…”

Section: Methodsmentioning

confidence: 99%

“…Kwiatkowski et al (2013) generates KB agnostic intermediary CCG parses of questions which are grounded afterwards given a KB. Bordes et al (2014) uses a vector space embedding approach to measure the semantic similarity between question and answers. Yao and Van Durme (2014), Bast and Haussmann (2015) and exploit a graph centric approach where a grounded subgraph query is generated from question and then executed against a KB.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Improving Semantic Parsing via Answer Type Inference

Yavuz¹,

Gür²,

Su³

et al. 2016

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

In this work, we show the possibility of inferring the answer type before solving a factoid question and leveraging the type information to improve semantic parsing. By replacing the topic entity in a question with its type, we are able to generate an abstract form of the question, whose answer corresponds to the answer type of the original question. A bidirectional LSTM model is built to train over the abstract form of questions and infer their answer types. It is also observed that if we convert a question into a statement form, our LSTM model achieves better accuracy. Using the predicted type information to rerank the logical forms returned by AgendaIL, one of the leading semantic parsers, we are able to improve the F1-score from 49.7% to 52.6% on the WE-BQUESTIONS data.

show abstract

Section: Methodsmentioning

confidence: 99%

mentioning

confidence: 99%

Improving Semantic Parsing via Answer Type Inference

Yavuz¹,

Gür²,

Su³

et al. 2016

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…These questions are subsequently interpreted with respect to the specific knowledge graph at hand by mapping them to a triple pattern query, which can be issued to the knowledge graph, returning the desired answers to the user [3,4,5,8]. This new setting of large knowledge graphs presents an opportunity to tackle the question answering problem using new approaches.…”

Section: Related Workmentioning

confidence: 99%

Answering Who/When, What, How, Why through Constructing Data Graph, Information Graph, Knowledge Graph and Wisdom Graph

Shao

Duan

Sun

et al. 2017

International Conferences on Software Engineering and Knowledge Engineering

View full text Add to dashboard Cite

Abstract-Knowledge graphs have been widely adopted, in large part owing to their schema-less nature. It enables knowledge graphs to grow seamlessly and allows for new relationships and entities as needed. Natural language questions are the most intuitive way of formulating an information need. People can formulate questions to express their information needs. Natural language questions as a query language present an ideal compromise between keyword and structured querying. Questions can be used to express complex information needs that cannot be expressed as keywords without a significant loss in structure and semantics. Knowledge graph has abundant natural semantics and can contain various and more complete information. Its expression mechanism is closer to natural language. We propose to clarify the expression of knowledge graph as a whole. We use knowledge graph to solve the Five Ws problems respectively which are guided by interrogative words such as who/when, what, how and why. We also propose to specify knowledge graph in a progressive manner as four basic forms including data graph, information graph, knowledge graph and wisdom graph.

show abstract

“…Recent research has focused more on developing open domain systems (Kwiatkowski et al, 2013;Yao and Durme, 2014;Bordes et al, 2014), but there remains a need for specialized NLIs (Minock, 2005). One unique feature of our system is to help users to build a complete question by providing suggestions according to a partial question and a grammar.…”

Section: Related Workmentioning

confidence: 99%