공간 질의응답 시스템에서의 개체링킹과 딥러닝 모델을 활용한 멘션 탐지 = Entity Linking in GeoKBQA: Approaches to Mention Detection Using Deep Learning Models

다국어 초록 (Multilingual Abstract)

With the rapid advancements in artificial intelligence and natural language processing technologies, the significance of question-answering (QA) systems between humans and machines has been on the rise. In this context, Knowledge Base Question Answering (KBQA) systems play a pivotal role in addressing diverse information needs of users. A KBQA system provides answers to user queries based on information stored in a Knowledge Base (KB). The typical pipeline of a standard KBQA system involves processing the natural language input, performing Entity linking to identify and link entities to their corresponding entries in the Knowledge Base, and subsequently generating a Logical Form. However, an extended domain of KBQA, known as Geographic Information-Based Question Answering (GeoKBQA), transforms spatially related queries into Logical Forms without the process of Entity linking. This research proposes a novel methodology to incorporate Entity linking in GeoKBQA to overcome such limitations. For this endeavor, deep learning models namely BERT, RoBERTa, and ChatGPT were utilized to carry out the initial process of Entity linking, which is Mention Detection (MD). Specifically, the BERT and RoBERTa models were fine-tuned with the NLMAPS dataset, comprised of geographic queries, for the specialized MD task concerning geographic inquiries. Conversely, ChatGPT employed a Few-shot prompting approach to perform MD. Experimental results indicated that the fine-tuned BERT and RoBERTa models achieved an F1 score of 0.96 and 0.97, while the ChatGPT model recorded an F1 score of 0.99.

참고문헌 (Reference)

1 류법모, "지식베이스 질의응답을 위한 자연언어 질문의 시맨틱 프레임 구조 표현 방법 연구" 한국정보기술학회 18 (18): 39-48, 2020

2 Wei S., "entity linking meets deep learning : techniques and solutions" 35 : 2556-2578, 2021

3 Dsouza, A., "WorldKG: A world-scale geographic knowledge graph" 2021

4 Hamzei, E., "Translating place-related questions to GeoSPARQL queries" 2022

5 Jiaxin, S., "Transfernet: An effective and transparent framework for multi-hop question answering over relation graph" 4149-4158, 2021

6 Ouyang, L., "Training language models to follow instructions with human feedback"

7 Shu, Y., "Tiara: Multi-grained retrieval for robust question answering over large knowledge base" 2022

8 Punjani, D., "Template-based question answering over linked geospatial data" 2018

9 Cui, L., "Template-based named entity recognition using BART"

10 Jing, Z., "Subgraph retrieval enhanced model for multi-hop knowledge base question answering" 2022

11 Wen-tau, Y., "Semantic parsing via staged query graph generation: Question answering with knowledge base" 1321-1331, 2015

12 Liu, Y., "RoBERTa: A robustly optimized BERT pretraining approach"

13 Ye, X., "Rng-kbqa: Generation augmented iterative ranking for knowledge base question answering" 2022

14 Chen, S., "ReTraCk: A flexible and efficient framework for knowledge base question answering" 2021

15 Yunshi, L., "Query graph generation for answering multi-hop complex questions from knowledge bases" 969-974, 2020

16 Cabrio, E., "QAKiS: an open domain QA system based on relational patterns" 2012

17 Damonte, M., "Practical semantic parsing for spoken language understanding"

18 Chen L., "Neural symbolic machines: Learning semantic parsers on freebase with weak supervision"

19 Yu, J., "Neural mention detection" 2020

20 Xixin, H., "Logical form generation via multi-task learning for complex question answering over knowledge bases" 2022

21 Brown, T. B., "Language models are few-shot learners"

22 Sewon, M., "Knowledge guided text retrieval and reading for open domain question answering"

23 Lawrence, C., "Improving a neural semantic parser by counterfactual learning from human bandit feedback" 2018

24 Yu, M., "Improved neural relation detection for knowledge base question answering" 1 : 2017

25 Mai, G., "Geographic question answering : challenges, uniqueness, classification, and future directions" 2 : 1-21, 2021

26 Evgeniy, G., "Facc1: Freebase annotation of clueweb corpora, version 1"

27 Xixin H., "EDG-based question decomposition for complex question answering over knowledge bases" 128-145, 2021

28 Yunshi L., "Complex knowledge base question answering: A survey" abs/2108. 06688 : 2021

29 Rajarshi D., "Case-based reasoning for natural language queries over knowledge bases" 9594-9611, 2021

30 Huiqiang, J., "BoningKnife: Joint entity mention detection and typing for nested NER via prior boundary knowledge" abs/2107.09429 : 2021

31 Yu G, "Beyond I.I.D.: three levels of generalization for question answering on knowledge bases" Association for Computing Machinery 3477-3488, 2021

32 Devlin, J., "BERT : Pre-training of deep bidirectional transformers for language understanding" North American Chapter of the Association for Computational Linguistics 2019

33 Abujabal, A., "Automated template generation for question answering over knowledge graphs" 2017

34 Yang, J., "Analyzing geographic questions using embedding-based topic modeling" 12 (12): 52-, 2023

35 Zheng, C., "A boundary-aware neural model for nested named entity recognition" 357-366, 2019