تشخیص رابطه پرسش‌های فارسی با ترکیب روش‌های مستقیم و غیرمستقیم

چکیده مقاله :

در این مطالعه برای مسأله پاسخ به سؤال‌های فارسی با استفاده از داده‌های پیوندی، زیرمسأله تشخیص رابطه برای سؤال‌های تک‌رابطه‌ای به تفصیل بررسی شده است. در این سؤال‌ها، پاسخ از یک سه‌تایی به شکل <فاعل، گزاره‌نما، مفعول> استخراج می‌شود. این کار دارای دو مرحله اصلی می‌باشد: نگاشت نهاد و تشخیص رابطه. در مرحله اول، نهاد شناسایی‌شده در سؤال به یک فاعل یا مفعول از یک سه‌تایی نگاشت شده و در مرحله دوم یک گزاره‌نما برای رابطه معنایی موجود در سؤال انتخاب می‌شود. در اکثر روش‌های موجود پس از نگاشت نهاد، همه رابطه‌های آن نهاد در پایگاه دانش به عنوان رابطه‌های نامزد در مرحله تشخیص رابطه در نظر گرفته شده و در نهایت یکی از آنها انتخاب می‌شود. در این روش‌ها اگر خطایی در مرحله نگاشت نهاد وجود داشته باشد به مرحله بعد منتشر شده و تشخیص رابطه به درستی انجام نمی‌شود. در این مطالعه برای رفع این وابستگی از ساختار سلسله‌مراتبی رابطه‌ها به منظور استخراج مستقیم رابطه‌ سؤال بهره گرفته می‌شود. دقت روش پیشنهادی در زبان فارسی برای تشخیص مستقیم رابطه 72% و برای انتخاب بهترین رابطه نامزد (غیرمستقیم) 90% می‌باشد. این دقت با ترکیب دو روش مستقیم و غیرمستقیم به 94% افزایش پیدا کرده است

چکیده انگلیسی :

In this study, for the problem of answering Persian questions using linked data, the sub-problem of relation detection for single-relation questions has been investigated in detail. In these questions, the answer is extracted from a triple in the form of <subject, predicate, object>. This process has two main steps: entity linking and relation detection. In the first step, the entity identified in the question is mapped to a subject or object of a triple, and in the second step, a predicate is selected for the semantic relation in the question. In most existing methods, after entity linking, all relations of that entity in the knowledge base are considered as candidate relations, and finally one of them is selected as the final relation. In these methods, if there is an error in the entity linking step, it is propagated to the relation detection step. In this study, to solve this dependency, the hierarchical structure of relations is used in order to directly extract the relation of the question. The accuracy of the proposed method in Persian is 72% for direct relation detection and 90% for selecting the best candidate relation (indirect). The accuracy has increased to 94% by combining direct and indirect methods.

 

منابع و مأخذ:

[1] S. Auer, et al., "DBedpia: a nucleus for a web of open data," in Proc. 6th Int. Semantic Web Con. and the 2nd Asian Semantic Web Conf., pp. 722-735, Busan, Sourh Korea,11-15 Nov. 2007.
[2] K. Bollacker, C. Evans, P. Paritosh, T. Sturge, and J. Taylor, "Freebase: a collaboratively created graph database for structuring human knowledge," in Proc. of the 2008 ACM SIGMOD Int. Conf. on Management of Data, pp. 1247-1250, Vancouver, BC, Canada, 10-12 Jun. 2008.
[3] M. Vegupatti, et al., "Simple question answering over a domain-specific knowledge graph using BERT by transfer learning," in Proc. 28th Irish Conf. on Artificial Intelligence and Cognitive Science, pp. 289-300, Dublin, Ireland, 7-8 Dec. 2020.
[4] H. Cui, T. Peng, L. Feng, T. Bao, and L. Liu, "Simple question answering over knowledge graph enhanced by question pattern classification," Knowl. Inf. Syst., vol. 63, no. 10, pp. 2741-2761, 2021.
[5] ش. بستان، ع. م. زارع بیدکی و م. ر. پژوهان، "بهبود رتبه‌بندی با استفاده از BERT،" نشریه مهندسی برق و مهندسی کامپیوتر ایران، ب- مهندسی کامپیوتر، سال 22، شماره 1، صص. 29-21، بهار 1403.
[6] A. S. Shamsabadi, R. Ramezani, H. K. Farsani, and M. Nematbakhsh, "Direct relation detection for knowledge-based question answering," Expert Syst. Appl., vol. 211, Article ID: 118678, 2023.
[7] D. Golub and X. He, "Character-level question answering with attention," in Proc. EMNLP 2016-Conf. on Empirical Methods in Natural Language Processing, pp. 1598-1607, Austin, Tx, USA, 1-4 Nov. 2016.
[8] W. Yin, M. Yu, B. Xiang, B. Zhou, and H. Schütze, "Simple question answering by attentive convolutional neural network," in Proc. 26th Int. Conf. on Computational Linguistics, Technical Papers, pp. 1746-1756, Osaka, Japan, 11-16 Dec. 2016.
[9] D. Lukovnikov, A. Fischer, J. Lehmann, and S. Auer, "Neural network-based question answering over knowledge graphs on word and character level," in Proc. 26th Int. World Wide Web Conf., pp. 1211-1220, Perth, Australia, 3-7 Apr. 2017.
[10] S. Mohammed, P. Shi, and J. Lin, "Strong baselines for simple question answering over knowledge graphs with and without neural networks," in Proc. of the Conf. of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, vol. 2, pp. 291-296, New Orleans, LA, USA, 1-6 Jun. 2018.
[11] G. Melis, C. Dyer, and P. Blunsom, "On the state of the art of evaluation in neural language models," in Proc. 6th Int. Conf. on Learning Representations, 10 pp., Vancouver, Canada, 30 Apr.-3 May 2018.
[12] A. Vaswani, et al., "Attention is all you need," Adv. Neural Inf. Process. Syst., vol. 30, pp. 5998-6008, 2017.
[13] M. Yu, et al., "Improved neural relation detection for knowledge base question answering," in Proc. 55th Annu. Meeting Assoc. Comput. Linguistics, pp. 571-581, Vancouver, Canada, 30 Jul.-4 Aug. 2017..
[14] H. Zhang, et al., "An attention-based word-level interaction model for knowledge base relation detection," IEEE Access, vol. 6, pp. 75429-75441, 2018..
[15] R. Z. Wang, Z. H. Ling, and Y. Hu, "Knowledge base question answering with attentive pooling for question representation," IEEE Access, vol. 7, pp. 46773-467842019..
[16] Y. Deng, et al., "Multi-task learning with multi-view attention for answer selection and knowledge base question answering," in Proc. 33rd AAAI Conf. on Artificial Intelligence, 31st Innovative Applications of Artificial Intelligence Conf., and the 9th AAAI Symposium on Educational Advances in Artificial Intelligence, pp. 6318-6325, Honolulu, HI, USA, 27 Jun.-1 Feb. 2019..
[17] Y. Chen and H. Li, "DAM: transformer-based relation detection for question answering over knowledge base," Knowledge-Based Syst., vol. 201-202, Article ID: pp. 106077, 2020..
[18] G. Maheshwari, et al., "Learning to rank query graphs for complex question answering over knowledge graphs," in Proc. Int. Semantic Web Conf., pp. 487-504, Auckland, New Zealand, 26-30 Oct.. 2019..
[19] "hazm 0.5.2." https://pypi.org/project/hazm/0.1/ .
[20] K. Taghva, R. Beckley, and M. Sadeh, "A stemming algorithm for the Farsi language," in Int. Conf. Inf. Technol. Coding Comput., vol. 2, pp. 158-162, Las Vegas, NV, USA, 4-6 Apr. 2005..
[21] Z. Mousavi and H. Faili, "Developing the persian wordnet of verbs using supervised learning," Trans. Asian Low-Resource Lang. Inf. Process., vol. 20, no. [22] E. Sherkat and M. Farhoodi, "A hybrid approach for question classification in Persian automatic question answering systems," in Proc. 4th Int. Conf. Comput. Knowl. Eng., pp. 279-284, Mashhad, Iran, 29-30 Oct. 2014..
[23] M. Razzaghnoori, H. Sajedi, and I. K. Jazani, "Question classification in Persian using word vectors and frequencies," Cogn. Syst. Res., vol. 47, pp. 16-27, Jan. 2018..
[24] F. Ahmadi and H. Moradi, "A hybrid method for Persian named entity recognition," in Proc. 7th Conf. on Information and Knowledge Technology, 7 pp., Urmia, Iran, 26-28 May 2015..
[25] H. Veisi and H. F. Shandi, "A Persian medical question answering system," Int. J. Artif. Intell. Tools, vol. 29, no. 6, Article ID: 2050019, 2020..
[26] A. Kazemi, J, Mozafari, and M. A. Nematbakhsh, "PersianQuAD: The native question answering dataset for the Persian language," IEEE Access, vol. 10, 26045-26057, 2022..
[27] R. Etezadi and M. Shamsfard, A Knowledge-Based Approach for Answering Complex Questions in Persian, arXiv Prepr. arXiv2107.02040, 2021..
[28] R. Etezadi and M. Shamsfard, "PeCoQ: a dataset for persian complex question answering over knowledge graph," in Proc. 11th Int. Conf. Inf. Knowl. Technol. pp. 102-106, Tehran, Iran, 22-23 Dec. 2020..
[29] F. Shirmardi, S. M. H. Hosseini, and S. Momtazi, "FarsWikiKG: an automatically constructed knowledge graph for Persian," Int. J. Web Res., vol. 4, no. 2, pp. 25-30, Dec. 2021..
[3‌ش م. ا. شناسا و ب. مینایی بیدگلی، "کاربست انواع جانمایی کلمات پیش‌آموزش داده‌شده در مدل‌های یادگیری عمیق برای تولید عنوان از متون فارسی،" نشریه مهندسی برق و مهندسی کامپیوتر ایران ب- مهندسی کامپیوتر، سال 22، شماره 1، صص. 38-30، بهار 1403..
[31] A. Bordes, N. Usunier, S. Chopra, and J. Weston, "Large-scale simple question answering with memory networks," CoRR, vol. abs/1506.0, [Online]. Available: http://arxiv.org/abs/1506.02075, 2015..
[32] P. Qi, T. Dozat, Y. Zhang, and C. D. Manning, "Universal dependency parsing from scratch," in Proc. of the {CoNLL} 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pp. 160-170, Brussels, Belgium, 31 Oct.-1 Nov. 2018..
[33] ش. بستان، ع. م. زارع بیدکی و م. ر. پژوهان، "درون‌سازی معنایی واژه‌ها با استفاده از BERT روی وب فارسی،" نشریه مهندسی برق و مهندسی کامپیوتر ایران، ب- مهندسی کامپیوتر، سال 21، شماره 2، صص. 100-89، تابستان 1402..
[34] -, "SimpleQuestions-EntityLinking," https://github.com/Gorov/SimpleQuestions-EntityLinking .
[35] م. حسین‌زاده اقدم، م. آنالویی و ج. تنها، "ارائه روشی جدید بر مبنای تجزیه ماتریس غیرمنفی برای کاهش ابعاد،" نشریه مهندسی برق و مهندسی کامپیوتر ایران، ب- مهندسی کامپیوتر، سال 20، شماره 2، صص. 172-164، تابستان 1401..
[36] W. T. Yih, M. W. Chang, X. He, and J. Gao, "Semantic parsing via staged query graph generation: question answering with knowledge base," in Proc. 53rd Annual Meeting of the Association for Computational Linguistics and the 7th Int. Joint Conf. on Natural Language Processing of the Asian Federation of Natural Language Processing, Proc. of the Conf., vol. 1, pp. 1321-1331, Beijing, China, 26-31 Jul. 2015..