Open Domain Answer Selection Model Fusing Double Matching-Focus

doi:10.19678/j.issn.1000-3428.0063516

Abstract

Abstract: Open-domain answer selection model is an important part of Question Answering (QA) system, which can find the answer that best matches the question by scoring different candidate answers for the same question.Current research on answer selection models in open-domain QA systems rarely focus on the fusion of word-level and sentence-level, which has resulted in lack of contextual semantic connections in matching relationships, or the loss of grammatical and semantic details of individual words.According to the adjacent similarity principle, this paper proposes an answer selection model fusing double matching-focus.Firstly, according to the characteristics of multi-sentence association in question and answer tasks, a word embedding method is designed that can embed the question and answer relationship and question and answer semantic relationship of words into word vectors, and uses word vectors to directly calculate the cosine similarity of word pairs to obtain word-level matching focus.then, sentence-level matching focus is extracted through the encoder-decoder model with added attention mechanism.finally, two focus distribution matrices are aligned with the question's words, and the relative distance between the focuses is obtained by merging word-level and sentence-level matching question-answer relevance score.Experimental results on two public QA datasets, Wiki-QA and TREC-QA, show that the mean Average Precision(mAP) index of the proposed answer selection model was 0.080 1 and 0.057 1, whilst the Mean Reciprocal Rank(MRR) index was improved by 0.017 6 and 0.006 6 compared to the multi-hop attention model and the hierarchical ranking model, respectively.

Key words: Question Answering(QA) system, adjacent similarity, matching-focus, word embedding vector, translation model

摘要： 开放域答案选择模型通过对同一问题的不同候选答案打分，寻找与问题最匹配的答案，是问答（QA）系统的重要组成部分。现有开放域QA系统中的答案选择模型较少关注词级与句子级的融合，导致在匹配关系上缺乏上下文的语义联系，或损失个别单词在语法语义上的细节信息。基于相邻相似原理，提出一种融合双匹配焦点的答案选择模型。根据问答任务多语句关联的特点，设计一种可以将词语的问答承接关系和问答语义关系嵌入进词向量的词嵌入方式，并利用该词向量直接计算词对的余弦相似度，得到词级匹配焦点。通过引入注意力机制的Encoder-Decoder模型提取句子级词对匹配焦点，以问题为基准对齐两个焦点分布矩阵，并使用焦点间的相对距离融合词级与句子级匹配矩阵，获得问题与答案的相关性得分。在Wiki-QA、TREC-QA两个公开问答数据集上的实验结果表明，该模型与多跳注意力模型、层级排序模型相比，平均准确率均值分别提高0.080 1和0.057 1，平均倒数排名分别提高0.017 6和0.006 6。

关键词: 问答系统, 相邻相似, 匹配焦点, 词嵌入向量, 翻译模型

CLC Number:

TP391

HE Junfei, ZHANG Huibing, HU Xiaoli. Open Domain Answer Selection Model Fusing Double Matching-Focus[J]. Computer Engineering, 2023, 49(1): 303-310.

何俊飞, 张会兵, 胡晓丽. 双匹配焦点融合的开放域答案选择模型[J]. 计算机工程, 2023, 49(1): 303-310.

/ Recommend / Download Citations

URL: https://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0063516

https://www.ecice06.com/EN/Y2023/V49/I1/303

Figures/Tables 10

References

[1] FANG H Y, WU F, ZHAO Z, et al.Community-based question answering via heterogeneous social network learning[C]//Proceedings of the 30th AAAI Conference on Artificial Intelligence.[S.l.]:AAAI Press, 2016:122-128.
[2] AMANCIO L, DORNELES C F, DALIP D H.Recency and quality-based ranking question in CQAs:a stack overflow case study[J].Information Processing&Management, 2021, 58(4):52-63.
[3] 朱宗奎, 张鹏举, 贾永辉, 等.基于多标签策略的中文知识图谱问答系统研究[J].计算机工程, 2021, 47(2):103-110, 117. ZHU Z K, ZHANG P J, JIA Y H, et al.Study of Chinese knowledge base question answering system based on multi-label strategy[J].Computer Engineering, 2021, 47(2):103-110, 117.(in Chinese)
[4] YU L, HERMANN K, BLUNSOM P, et al.Deep learning for answer sentence selection[EB/OL].[2021-09-03].https://www.semanticscholar.org/paper/Deep-Learning-for-Answer-Sentence-Selection-Yu-Hermann/4cfad7889dc12825309325cd4b4f3febed424e36.
[5] WANG D, NYBERG E.A long short-term memory model for answer sentence selection in question answering[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2:Short Papers).Stroudsburg, USA:Association for Computational Linguistics, 2015:707-712.
[6] HE H, GIMPEL K, LIN J.Multi-perspective sentence similarity modeling with convolutional neural networks[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing. Stroudsburg, USA:Association for Computational Linguistics, 2015:1576-1586.
[7] HU B, LU Z, LI H, et al.Convolutional neural network architectures for matching natural language sentences[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems.New York, USA:ACM Press, 2014:2042-2050.
[8] WANG S, JIANG J.A compare-aggregate model for matching text sequences[EB/OL].[2021-09-03].https://www.zhuanzhi.ai/paper/6ab06cb2717944c4a925e7b116875e15.
[9] BAHDANAU D, CHO K H, BENGIO Y.Neural machine translation by jointly learning to align and translate[EB/OL].[2021-09-03].https://www.zhuanzhi.ai/paper/6ab06cb2717944c4a925e7b116875e15.
[10] SUTSKEVER I, VINYALS O, LE Q V.Sequence to sequence learning with neural networks[EB/OL].[2021-09-03].http://static.googleusercontent.com/media/research.google.com/zh-CN//pubs/archive/43155.pdf.
[11] HERMANN K M, KOCISKY T, GREFENSTETTE E, et al.Teaching machines to read and comprehend[J].Advances in Neural Information Processing Systems, 2015, 28:1693-1701.
[12] ROCKTÄSCHEL T, GREFENSTETTE E, HERMANN K, et al.Reasoning about entailment with neural attention[EB/OL].[2021-09-03].https://www.semanticscholar.org/paper/Reasoning-about-Entailment-with-Neural-Attention-Rockt%C3%A4schel-Grefenstette/2846e83d405cbe3bf2f0f3b5f635dd8b3c680c45.
[13] VASIN P, ROTH D, YIH W T.Mapping dependencies trees:an application to question answering[EB/OL].[2021-09-03].https://www.docin.com/p-957607371.html.
[14] YIH W T, CHANG M W, MEEK C, et al.Question answering using enhanced lexical semantic models[EB/OL].[2021-09-03].https://www.semanticscholar.org/paper/Question-Answering-Using-Enhanced-Lexical-Semantic-Yih-Chang/f1aa6df7f18f9cb7d6b6c5c190aeade47b450656.
[15] SHAH C, POMERANTZ J.Evaluating and predicting answer quality in community QA[C]//Proceedings of the 33rd International Conference on Research and Development in Information Retrieval.New York, USA:ACM Press, 2010:411-418.
[16] TRAN N K, NIEDEREÉE C.Multihop attention networks for question answer matching[C]//Proceedings of the 41st International ACM SIGIR Conference on Research & Development in Information Retrieval.New York, USA:ACM Press, 2018:325-334.
[17] 唐竑轩, 武恺莉, 朱朦朦, 等.基于双向注意力机制的多文档神经阅读理解[J].计算机工程, 2020, 46(12):43-51. TANG H X, WU K L, ZHU M M, et al.Multi-document neural reading comprehension based on bi-directional attention mechanism[J].Computer Engineering, 2020, 46(12):43-51.(in Chinese)
[18] YOON S, DERNONCOURT F, KIM D S, et al.A compare-aggregate model with latent clustering for answer selection[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management.New York, USA:ACM Press, 2019:2093-2096.
[19] LI L S, ZHOU A Q, ZHANG B B, et al.Multiple fragment-level interactive networks for answer selection[J].Neurocomputing, 2020, 402:80-88.
[20] GAO H, HU M, CHENG R, et al.Hierarchical ranking for answer selection[EB/OL].[2021-09-03].https://arxiv.org/abs/2102.00677.
[21] MIKOLOV T, CHEN K, CORRADO G, et al.Efficient estimation of word representations in vector space[EB/OL].[2021-09-03].https://www.semanticscholar.org/paper/Efficient-Estimation-of-Word-Representations-in-Mikolov-Chen/330da625c15427c6e42ccfa3b747fb29e5835bf0.
[22] PENNINGTON J, SOCHER R, MANNING C.Glove:global vectors for word representation[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing.Stroudsburg, USA:Association for Computational Linguistics, 2014:1532-1543.
[23] JACOB D, CHANG M W, KENTON L, et al.BERT:pre-training of deep bidirectional transformers for language understanding.[EB/OL].[2021-09-03].https://arxiv.org/abs/1810.04805.
[24] PETERS M, NEUMANN M, IYYER M, et al.Deep contextualized word representations[C]//Proceedings of Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies, Volume 1(Long Papers).Stroudsburg, USA:Association for Computational Linguistics, 2018:2227-2237.
[25] LASKAR M T R, HUANG X, HOQUE E.Contextualized embeddings based transformer encoder for sentence similarity modeling in answer selection task[EB/OL].[2021-09-03].https://www.researchgate.net/publication/341231839_Contextualized_Embeddings_based_Transformer_Encoder_for_Sentence_Similarity_Modeling_in_Answer_Selection_Task.
[26] WANG M, SMITH N A, MITAMURA T.What is the Jeopardy model?a quasi-synchronous grammar for QA[EB/OL].[2021-09-03].https://aclanthology.org/D07-1003.pdf.
[27] YANG Y, YIH W T, MEEK C.WikiQA:a challenge dataset for open-domain question answering[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing.Stroudsburg, USA:Association for Computational Linguistics, 2015:2013-2018.

Please choose a citation manager

Content to export