Open-domain Factoid Question Answering via Knowledge Graph Search

Aghaebrahimian, Ahmad; Jurčíček, Filip

doi:10.18653/v1/w16-0104

Cited by 16 publications

(6 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, QA has been largely benefited from the development of Deep Neural Network (DNN) architectures largely in the form of Convolution Neural Networks (CNN) (LeCun et al, 1998) or Recurrent Neural Networks (RNN) (Elman, 1990). QA systems based on semantic parsing (Clarke et al, 2010;Kwiatkowski et al, 2010), IR-based systems (Yao and Durme, 2014), cloze-type (Kadlec et al, 2016;Hermann et al, 2015), factoid (Aghaebrahimian and Jurčíček, 2016b; and non-factoid systems (Aghaebrahimian, 2017a;Rajpurkar et al, 2016) are some of the QA variants that have been improved by DNNs. Among all of these varieties, factoid and non-factoid are two most widely studied branches of QA systems.…”

Section: Related Workmentioning

confidence: 99%

Linguistically-Based Deep Unstructured Question Answering

Aghaebrahimian

2018

Proceedings of the 22nd Conference on Computational Natural Language Learning

Self Cite

View full text Add to dashboard Cite

In this paper, we propose a new linguisticallybased approach to answering non-factoid open-domain questions from unstructured data. First, we elaborate on an architecture for textual encoding based on which we introduce a deep end-to-end neural model. This architecture benefits from a bilateral attention mechanism which helps the model to focus on a question and the answer sentence at the same time for phrasal answer extraction. Second, we feed the output of a constituency parser into the model directly and integrate linguistic constituents into the network to help it concentrate on chunks of an answer rather than on its single words for generating more natural output. By optimizing this architecture, we managed to obtain near-to-human-performance results and competitive to a state-of-the-art system on SQuAD and MS-MARCO datasets respectively.

show abstract

Section: Related Workmentioning

confidence: 99%

Linguistically-Based Deep Unstructured Question Answering

Aghaebrahimian

2018

Proceedings of the 22nd Conference on Computational Natural Language Learning

Self Cite

View full text Add to dashboard Cite

show abstract

“…We look up the extracted fragments in Wikidata by comparing them to labels of the Wikidata items. Following the approach in [1], we sort the retrieved list of items by the combination of the Levenshtein distance between the fragment and the item label and the integer part of the item ID. We select the top candidate for each fragment as the final linking.…”

Section: Entity Linkingmentioning

confidence: 99%

“…We tokenize the relation labels and use them as input to the same CNN-based encoder to produce a semantic vector for each relation. 1 To get a single vector for the whole representation, we apply another max pooling operation on the set of the relation vectors. The final semantic vector for a candidate representation encodes the most prominent features of the relations that it contains.…”

Section: Iterative Representation Generationmentioning

confidence: 99%

End-to-End Representation Learning for Question Answering with Weak Supervision

Sorokin

Gurevych

2017

Semantic Web Challenges

View full text Add to dashboard Cite

In this paper we present a factoid question answering system for participation in Task 4 of the QALD-7 shared task. Our system is an end-to-end neural architecture for learning a semantic representation of the input question. It iteratively generates representations and uses a convolutional neural network (CNN) model to score them at each step. We take the semantic representation with the highest final score and execute it against Wikidata to retrieve the answers. We show on the Task 4 data set that our system is able to successfully generalize to new data.

show abstract

“…This process requires detailed knowledge to establish a high coverage and e cient question answering system [5,6]. In such systems, multiple knowledge formats from a large variety of natural language sources, such as textbooks, encyclopedias, newspapers, and literary works, have to be processed to provide a knowledge base [7].…”

Section: Introductionmentioning

confidence: 99%

A Statistical Approach for Knowledge Discovery: Bootstrapped Analysis of Language Models for Knowledge base Population from Unstructured Text

Momtazi

Moradiannasab

2018

Scientia Iranica

View full text Add to dashboard Cite

Abstract. This paper proposes a novel approach to knowledge discovery from textual data. The generated knowledge base can be used as one of the main components in the cognitive process of question answering systems. The proposed model automatically extracts relations between named entities in Persian. Our proposed model is a bootstrapping approach based on n-gram (a contiguous sequence of n items from a given sequence of text or speech) model to nd the representative textual patterns of relations as n-grams in order to extract new knowledge about given named entities. The main motivation of this work is the characteristic of the sentence structure in Persian which, in comparison to English sentences, is in subject-object-verb format. The proposed approach is a purely statistical one, and no background knowledge of the target language is required. This makes our method applicable to any open domain relation extraction task. However, as for our test-bed, the domain of biographical data of international poets and scientists is considered herein to build a knowledge base about them. Qualitative evaluations based on human assessment represent the evidence of the e cacy of our method.

show abstract

Open-domain Factoid Question Answering via Knowledge Graph Search

Cited by 16 publications

References 9 publications

Linguistically-Based Deep Unstructured Question Answering

Linguistically-Based Deep Unstructured Question Answering

End-to-End Representation Learning for Question Answering with Weak Supervision

A Statistical Approach for Knowledge Discovery: Bootstrapped Analysis of Language Models for Knowledge base Population from Unstructured Text

Contact Info

Product

Resources

About