SCIENCE CHINA Information Sciences, Volume 60, Issue 11: 110103(2017) https://doi.org/10.1007/s11432-017-9136-x

## Topic enhanced deep structured semantic models for knowledge base question answering

• AcceptedJun 20, 2017
• PublishedSep 21, 2017
Share
Rating

### Abstract

Knowledge Base Question Answering (KBQA) is a hot research topic in natural language processing (NLP). The most challenging problem in KBQA is how to understand the semantic information of natural language questions and how to bridge the semantic gap between the natural language questions and the structured fact triples in knowledge base. This paper focuses on simple questions which can be answered by a single fact triple in knowledge base. We propose a topic enhanced deep structured semantic model for KBQA. The proposed method considers the task of KBQA as a matching problem between questions and the subjects and predicates in knowledge base. And the proposed model consists of two stages to match the subjects and predicates, respectively. In the first stage, we propose a Convolutional based Topic Entity Extraction Model (CTEEM) to extract topic entities mentioned in questions. With the extracted entities, we can retrieve the relevant candidate fact triples from knowledge base and obviously decrease the amount of noising candidates. In the second stage, we employ Deep Structured Semantic Models (DSSMs) to compute the semantic relevant score between questions and predicates in the candidates. And we combine the semantic level and the lexical level scores to rank the candidates. We evaluate the proposed method on KBQA dataset released by NLPCC-ICCPOL 2016. The experimental results show that our proposed method achieves the third place among the 21 submitted systems. Furthermore, we also extend the DSSM by using BiLSTM and integrate a convolutional structure on the top of BiLSTM layers. Our experimental results show that the extension models can further improve the performance.

### Acknowledgment

This work was supported by National Natural Science Foundation of China (Grant Nos. 61573163, 71571084), Fundamental Research Funds for the Central Universities (Grant No. CCNU16A02024), and Wuhan Youth Science and Technology Plan.

### References

[1] Bollacker K, Evans C, Paritosh P, et al. Freebase: a collaboratively created graph database for structuring human knowledge. In: Proceedings of the 2008 ACM SIGMOD International Conference on Management of Data, Vancouver, 2008. 1247--1250. Google Scholar

[2] Auer S, Bizer C, Kobilarov G, et al. Dbpedia: a nucleus for a web of open data. In: Proceedings of the 6th International the Semantic Web and 2nd Asian Conference on Asian Semantic Web, Busan, 2007. 722--735. Google Scholar

[3] Berant J, Chou A, Frostig R, et al. Semantic parsing on freebase from question-answer pairs. Proc EMNLP, 2013, 2: 1533--1544. Google Scholar

[4] Berant J, Liang P. Semantic parsing via paraphrasing. In: Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, Stroudsburg, 2014. 1415--1425. Google Scholar

[5] Bordes A, Usunier N, Chopra S, et al. Large-scale simple question answering with memory networks. Comput Sci, 2015,. arXiv Google Scholar

[6] Bordes A, Chopra S, Weston J. Question answering with subgraph embeddings. Comput Sci, 2014,. arXiv Google Scholar

[7] Bordes A, Weston J, Usunier N. Open question answering with weakly supervised embedding models. In: Proceedings of Joint European Conference on Machine Learning and Knowledge Discovery in Databases, Nancy, 2014. 165--180. Google Scholar

[8] Yao X, Durme B V. Information extraction over structured data: question answering with freebase. In: Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, Stroudsburg, 2014. 956--966. Google Scholar

[9] Shen Y, He X, Gao J, et al. Learning semantic representations using convolutional neural networks for web search. In: Proceedings of the 23rd International Conference on World Wide Web, Seoul, 2014. 373--374. Google Scholar

[10] Zettlemoyer L S, Collins M. Learning to map sentences to logical form: structured classification with probabilistic categorial grammars. In: Proceedings of the 21st Conference on Uncertainty in Artificial Intelligence, Edinburgh, 2012. Google Scholar

[11] Kwiatkowski T, Zettlemoyer L, Goldwater S, et al. Inducing probabilistic CCG grammars from logical form with higher-order unification. In: Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, Cambridge, 2010. 1223--1233. Google Scholar

[12] Liang P, Jordan M I, Klein D. Learning dependency-based compositional semantics. In: Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies-Volume 1, Stroudsburg, 2011. 590--599. Google Scholar

[13] Cai Q, Yates A. Large-scale semantic parsing via schema matching and lexicon extension. In: Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, Sofia, 2013. 423--433. Google Scholar

[14] Lai Y, Lin Y, Chen J, et al. Open domain question answering system based on knowledge base. In: Proceedings of the 24th International Conference on Computer Processing of Oriental Languages, Kunming, 2016. 722--733. Google Scholar

[15] Wang L, Zhang Y, Liu T. A deep learning approach for question answering over knowledge base. In: Proceedings of the 24th International Conference on Computer Processing of Oriental Languages, Kunming, 2016. 885--892. Google Scholar

[16] Yang F, Gan L, Li A, et al. Combining deep learning with information retrieval for question answering. In: Proceedings of the 24th International Conference on Computer Processing of Oriental Languages, Kunming, 2016. 917--925. Google Scholar

[17] Xie Z, Zeng Z, Zhou G, et al. Knowledge base question answering based on deep learning models. In: Proceedings of the 24th International Conference on Computer Processing of Oriental Languages, Kunming, 2016. 300--311. Google Scholar

[18] Pennington J, Socher R, Manning C D. Glove: global vectors for word representation. Proc EMNLP, 2014, 14: 1532--1543. Google Scholar

[19] Mikolov T, Sutskever I, Chen K, et al. Distributed representations of words and phrases and their compositionality. Adv Neural Inform Process Syst, 2013, 26: 3111--3119. Google Scholar

[20] Mikolov T, Chen K, Corrado G, et al. Efficient estimation of word representations in vector space. Comput Sci, 2013,. arXiv Google Scholar

[21] Zhou G Y, He T T, Zhao J, et al. Learning continuous word embedding with metadata for question retrieval in community question answering. In: Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, Beijing, 2015. 250--259. Google Scholar

[22] Zhou G, Huang J X. Modeling and Learning Distributed Word Representation with Metadata for Question Retrieval. IEEE Trans Knowl Data Eng, 2017, 29: 1226-1239 CrossRef Google Scholar

[23] Huang P S, He X, Gao J, et al. Learning deep structured semantic models for web search using clickthrough data. In: Proceedings of the 22nd ACM International Conference on Information & Knowledge Management, San Francisco, 2013. 2333--2338. Google Scholar

[24] Shen Y, He X, Gao J, et al. A latent semantic model with convolutional-pooling structure for information retrieval. In: Proceedings of the 23rd ACM International Conference on Information and Knowledge Management, Shanghai, 2014. 101--110. Google Scholar

[25] Bahdanau D, Cho K, Bengio Y. Neural machine translation by jointly learning to align and translate. Comput Sci, 2014,. arXiv Google Scholar

[26] Rush A M, Chopra S, Weston J. A neural attention model for abstractive sentence summarization. Proc EMNLP, 2015. Google Scholar

[27] Yih W, Chang M W, He X, et al. Semantic parsing via staged query graph generation: question answering with knowledge base. In: Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics, Beijing, 2015. Google Scholar

[28] Yih W T, He X, Meek C. Semantic parsing for single-relation question answering. In: Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, Baltimore, 2014. 643--648. Google Scholar

[29] Dong L, Wei F, Zhou M, et al. Question answering over freebase with multi-column convolutional neural networks. In: Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics, Beijing, 2015. 260--269. Google Scholar

[30] Zhang Y, Liu K, He S, et al. Question answering over knowledge base with neural attention combining global knowledge information. Comput Sci, 2016,. arXiv Google Scholar

[31] Jain S. Question answering over knowledge base using factual memory networks. In: Proceedings of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), San Diego, 2016. 109--115. Google Scholar

[32] Dai Z H, Li L, Xu W. Cfo: conditional focused neural question answering with large-scale knowledge bases. In: Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, Berlin, 2016. Google Scholar

[33] Lafferty J, McCallum A, Pereira F. Conditional random fields: probabilistic models for segmenting and labeling sequence data. In: Proceedings of the 18th International Conference on Machine Learning, San Francisco, 2001. 282--289. Google Scholar

[34] Palangi H, Deng L, Shen Y, et al. Semantic modelling with long-short-term memory for information retrieval. Comput Sci, 2014,. arXiv Google Scholar

[35] Hochreiter S, Schmidhuber J. Long Short-Term Memory. Neural Computation, 1997, 9: 1735-1780 CrossRef Google Scholar

[36] Gers F A, Schmidhuber J, Cummins F. Learning to Forget: Continual Prediction with LSTM. Neural Computation, 2000, 12: 2451-2471 CrossRef Google Scholar

[37] Gers F A, Schraudolph N N, Schmidhuber J. Learning precise timing with LSTM recurrent networks. J Mach Learn Res, 2002, 3: 115--143. Google Scholar

[38] Graves A, Mohamed A, Hinton G. Speech recognition with deep recurrent neural networks. In: Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing, Vancouver, 2013. 6645--6649. Google Scholar

[39] Duan N. Overview of the NLPCC-ICCPOL 2016 shared task: open domain chinese question answering. In: Proceedings of the 24th International Conference on Computer Processing of Oriental Languages, Kunming, 2016. 942--948. Google Scholar

[40] Kingma D, Ba J. Adam: a method for stochastic optimization. In: Proceedings of the 3rd International Conference for Learning Representations, San Diego, 2014. Google Scholar

• Figure 1

(Color online) The overview of the KBQA system framework.

• Figure 2

(Color online) The architecture of the CTEEM for an example question.

• Figure 3

(Color online) The architecture of the CDSSM used in this paper.

• Figure 4

(Color online) The architecture of BDSSM.

• Figure 5

(Color online) An example of the lexical level matching.

• Figure 6

(Color online) The average $F_1$ vs. $\omega$ curves of the three semantic matching methods.

• Table 1   Topic entities extracted by the CTEEM
 Question Topic entity 你$|$知道$|$雷锋$|$日记$|-|$拼音$|$版$|$的$|$副标题$|$是$|$什么$|$吗$|$? 雷锋日记–拼音版 你$|$知道$|$倭$|$叉$|$角$|$羚$|$这种$|$动物$|$是$|$什么$|$纲$|$的$|$吗$|$? 倭叉角羚 你$|$知道$|$拼搏$|$奥运$|$连连看$|$占$|$多$|$大$|$的$|$内存$|$吗$|$? 拼搏奥运连连看 奥图$|$码$|$pv$|$3225$|$这款$|$产品$|$是$|$干什么$|$用$|$的$|$啊$|$? 奥图码 pv 3225 恒大$|$金碧$|$天下$|$是$|$什么$|$样子$|$的$|$房子$|$啊$|$? 恒大金碧天下 上海$|$假日$|$之$|$星$|$酒店$|$($|$长宁店$|$)$|$在$|$哪$|$啊 $|$? 上海假日之星酒店 你$|$知道$|$资产$|$预计$|$未来$|$现金流量$|$是$|$什么$|$意思$|$吗$|$? 资产预计未来现金流量
• Table 2   The comparison of the candidates obtained by using the extracted topic entity and n-gram method
 CTEEM method n-gram method noalign Total number 6 3000 Candidates 飞廉状风毛菊$|||$别名$|||$飞廉状风毛菊 于中国$|||$别名$|||$于中国 飞廉状风毛菊$|||$中文名$|||$飞廉状风毛菊 于中国$|||$中文名$|||$于中国 飞廉状风毛菊$|||$门$|||$被子植物门 于中国$|||$国籍$|||$中国 飞廉状风毛菊$|||$国内分布$|||$陕西省, 甘肃省, 四川省 于中国$|||$职业$|||$高级工程师 飞廉状风毛菊$|||$生境$|||$疏林中 于中国$|||$毕业院校$|||$吉林大学 飞廉状风毛菊$|||$栽培$|||$非人工引种栽培 飞廉状风毛菊$|||$别名$|||$飞廉状风毛菊 飞廉状风毛菊$|||$中文名$|||$飞廉状风毛菊 飞廉状风毛菊$|||$门$|||$被子植物门 飞廉状风毛菊$|||$国内分布$|||$陕西省, 甘肃省, 四川省 飞廉状风毛菊$|||$生境$|||$疏林中 飞廉状风毛菊$|||$栽培$|||$非人工引种栽培 分布(汉语词汇)$|||$别名$|||$分布 分布(汉语词汇)$|||$中文名$|||$分布 $\cdots~\cdots$
• Table 3   The experimental results on the test dataset
 System $\mbox{Average}~F_1~(%)$ Baseline system (CDSSM) 52.47 Lai et al., 2016[14] 82.47 Compared methods Yang et al., 2016[16] 81.59 Wang et al., 2016[15] 79.14 Xie et al., 2016[17] 79.57 CTEEM+LMS 74.62 CTEEM+CDSSM 75.74 CTEEM+BDSSM 74.22 CTEEM+BCDSSM 76.18 Ours CTEEM+Combined-DSSM 77.89 CTEEM+CDSSM+LMS 81.89 CTEEM+BDSSM+LMS 81.63 CTEEM+BCDSSM+LMS 82.21 CTEEM+Combined-DSSM+LMS 82.43
• #### 7

Citations

• Altmetric

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有