Tibetan automated question-answering (Q&A) plays a significant role in the development of artificial intelligence interdisciplinary fields such as natural language processing, computational linguistics, knowledge engineering, and cross-language information processing. Q&A datasets are the cornerstone of studying automated Q&A, yet currently, the methods for Tibetan Q&A dataset construction are limited, and Tibetan Q&A open-source datasets are scarce. Therefore, this paper proposed a diversified approach for constructing a Tibetan Q&A dataset through an in-depth comparative analysis of publicly available datasets both domestically and internationally. First, it expands data sources by collecting and organizing diversified Q&A data from Tibetan websites and non-Tibetan datasets. Then, it enhances the diversification of Tibetan Q&A data by integrating Weibo dialogue data, domain-specific Q&A data, and machine reading comprehension-based Q&A data. Finally, through the comprehensive application of web scraping, BERT-based sentiment classification, translation of non-Tibetan Q&A data, and normalization and correction of Tibetan Q&A data, a diversified Tibetan Q&A dataset comprising 84,118 pairs is constructed. The experimental results demonstrate that the constructed dataset achieves 44.12% BLEU and 78.63% EmbAve on the Seq2Seq model, representing improvements of 0.96 and 1.42 percentage points over the baseline Qingyun datasets, for which the BLEU and EmbAve are 43.16% and 77.21%, respectively. This fully validates its effectiveness for the Q&A generation tasks.
GreenB F, WolfA K, ChomskyC, et al. Baseball: An automatic question-answerer[C]. Proceedings of the Western Joint Computer Conference, 1961:219-224.
[3]
WoodsW A. Semantics and quantification in natural language question answering[M]. Advances in computers. Elsevier, 1978, 17:1-87.
[4]
StolckeA, RiesK, CoccaroN, et al. Dialogue act modeling for automatic tagging and recognition of conversational speech[J]. Computational linguistics, 2000,26(3):339-373.
[5]
LendvaiP, GeertzenJ. Token-based chunking of turn-internal dialogue act sequences[C]. Proceedings of the 8th SIGDIAL Workshop on Discourse and Dialogue, 2007:174-181.
[6]
ZimmermannM. Joint segmentation and classification of dialog acts using conditional random fields[C]. Conference of the International Speech Communication Association(INTERSPEECH), 2009:864-867.
[7]
YanF, SilamuW, ChaiY, et al. OECA-Net: A co-attention network for visual question answering based on OCR scene text feature enhancement [J]. Multimedia Tools and Applications, 2024,83(3):7085-7096.
[8]
ZhangW, LiuQ. Convolutional neural network for substantiation of children's books, intelligent interactive communication and application analysis of voice question answering [J]. Entertainment Computing, 2024,48:100608.
SunY, XiaT. A hybrid network model for Tibetan question answering [J]. IEEE Access, 2019,7:52769-52777.
[11]
SunY, ChenC, XiaT, et al. QuGAN: quasi generative adversarial network for Tibetan question answering corpus generation [J]. IEEE Access, 2019,7:116247-116255.
[12]
SunY, ChenC, ChenA, et al. Tibetan question generation based on sequence to sequence model[J]. Computers, Materials & Continua, 2021,68(3).
[13]
LiuS, ChenC, SunY. Paragraph-level Tibetan question generation for machine reading comprehension[C]. 2022 International Conference on Asian Language Processing(IALP). IEEE, 2022:280-285.
[14]
SunY, LiuS, DanZ, et al. Question generation based on grammar knowledge and fine-grained classification[C]. Proceedings of the 29th International Conference on Computational Linguistics. 2022:6457-6467.
WangY, KeP, ZhengY, et al. A large-scale chinese short-text conversation dataset [J]. International conference natural language processing, 2020:91-103.
[18]
DevlinJ, ChangM W, LeeK, et al. Bert: Pre-training of deep bidirectional transformers for language understanding[C]. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human language Technologies, 2019: 4171-4186.
[19]
SutskeverI, VinyalsO, LeQ V. Sequence to sequence learning with neural networks[J]. Advances in Neural Information Processing Systems, 2014,27.
[20]
ChiranjeeviR. Chatbot for government schemes using SEQ2SEQ model[C]. 2024 Second International Conferenceon Advances in Information Technology (ICAIT). IEEE, 2024:1-6.