ABSTRACT
In recent years, as the field of new energy power has gradually become a research hotspot, there are more and more research results related to new energy power. This paper first proposes to Fine-tune the Chinese LLaMA large language model to realize the extraction of research questions and methods in new energy power results. The fine-tuning dataset is constructed by the combination of rule template and gpt-3.5 enhancement, which avoids the costly and time-consuming problem caused by manual construction. The fine-tuning method adopts LoRA high-efficiency fine-tuning to save computing resources; Then, F1 value is used as the evaluation index to compare the extraction effect of the model under different fine-tuning datasets. The results show that the model has a good extraction effect on the research questions and method terms when training the dataset constructed by the combination of rule template and gpt-3.5 enhancement. Finally, according to the extracted research question phrases, BTM(Biterm Topic Model) is used to study the distribution of topic words, and soft clustering of research question phrases is carried out according to the obtained topic words, so as to realize the correlation between the research results and professional terms, which provides the foundation for the future establishment of the knowledge graph and knowledge base of new energy power.
CCS CONCEPTS • Theory of computation • Theory and algorithms for application domains • Unsupervised learning and clustering
- YANG Wei, SUN Deyan, ZHANG Xiaohui, .Named entity recognition for intelligent answer system in power service[J]. Computer Engineering and Design,2019,40(12):3625-3630(in Chinese).Google Scholar
- YANG Q Y, JIANG J, FENG X Y, Named entity recognition of power substation knowledge based on transformer-BiLSTM-CRF network[C]//2020 International Conference on Smart Grids and Energy Systems (SGES).Perth:IEE,2020:952-956.Google Scholar
- XU Huifang, ZHANG Zhonghao, TAN Yuanpeng, .Research on entity recognition technology in power grid dispatching field[J].Electric Power Construction,2021,42(10):71-77(in Chinese).Google Scholar
- CHEN Peng, TAI Bin, SHI Ying, et al. Text Entity Extraction of Power Equipment Defects Based on BERT-BI-LSTM-CRF Algorithm. Power System Technology,2023,47(10):4367-4376.Google Scholar
- Vaswani A, Shazeer N, Parmar N, Attention is all you need[C/OL]//Guyon I, Luxburg U V, Bengio S, Advances in Neural Information Processing Systems: volume 30. Curran Associates, Inc.,2017. https://proceedings.neurips.cc/paper/2017/file/ 3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf.Google Scholar
- Devlin J, Chang M W, Lee K, 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, Volume 1 (Long and Short Papers). 2019: 4171-4186.Google Scholar
- Radford A, Wu J, Child R, Language models are unsupervised multitask learners[J]. OpenAI blog, 2019, 1(8):9.Google Scholar
- Touvron H, Lavril T, Izacard G, Llama: Open and efficient foundation language models[J].arXiv preprint arXiv:2302.13971, 2023Google Scholar
- Taori R, Gulrajani I, Zhang T, Stanford alpaca: An instruction-following llama model[J/OL]. GitHub repository, 2023. https://github.com/tatsu-lab/stanford_alpaca.Google Scholar
- Hu E J, yelong shen, Wallis P, LoRA: Low-rank adaptation of large language models[C/OL]//International Conference on LearninRepresentations. 2022. https://openreview.net/forum?id=nZeVKeeFYf9.Google Scholar
- Aghajanyan A, Zettlemoyer L, Gupta S. Intrinsic dimensionality explains the effectiveness of language model fine-tuning[J]. arXiv preprint arXiv:2012.13255, 2020.Google Scholar
- Brown T, Mann B, Ryder N, Language models are few-shot learners[J]. Advances in neural information processing systems, 2020, 33:1877-1901.Google Scholar
- Touvron H, Martin L, Stone K, Llama 2: Open foundation and fine-tuned chat models[J]. arXiv preprint arXiv:2307.09288, 2023Google Scholar
- Rasley J, Rajbhandari S, Ruwase O, Deepspeed: System optimizations enable training deep learning models with over 100 billion parameters[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2020: 3505-3506.Google Scholar
- Bommasani R, Liang P, Lee T. Holistic evaluation of language models[J]. Annals of the New York Academy of Sciences,2023.Google Scholar
- https://github.com/ymcui/Chinese-LLaMA-AlpacaGoogle Scholar
- FENG Bin, ZHANG Wenwen, TANG Xin, Power Equipment Defect Record Text Mining Based on BiLSTM-Attention Neural Network, Proceedings of the CSEE, 2020,40(S1): 1-10Google Scholar
- LI Yanxuan. Research and implementation of abstract-based paper classification and recommendation model. Beijing: Beijing University of Posts and Telecommunications, 2019.Google Scholar
- ZHAI Yujia, TIAN Jingwen, ZHAO Ban. Algorithm Term Extraction and Innovation Evolution Path Construction Based on BERT-BiLSTM-CRF Model. Information Science, 2022,40(4):71-78Google Scholar
- HAN Hongqi, XU Shuo, GUI Jie, Term Hierarchical Relation Extraction Method Based on Morphology Rule Template. Information Science, 2013,32(7):708-715.Google Scholar
- Xiaohui Yan, Jiafeng Guo, Yanyan Lan, A biterm topic model for short texts.Proceedings of the 22nd international conference on World Wide Web.May,2013.Pages 1445–1456.https://doi.org/10.1145/2488388.2488514Google ScholarDigital Library
- GAO Huiying, GONG Mengqiu, YU Sijia. Identification of Medical Service Quality Factors Based on COA-BTM Model. Journal of Beijing Institute of Technology, 2022,(11):1167-1174Google Scholar
- XU Feifei, CHEN Saihong, TIAN Yu. Hot topic detection based on BTM model and improved clustering algorithm. Computer Applications And Software, 2022,(5):283-290.Google Scholar
Recommendations
A graph-based topic extraction method enabling simple interactive customization
DocEng '13: Proceedings of the 2013 ACM symposium on Document engineeringIt is often desirable to identify the concepts that are present in a corpus. A popular way to deal with this objective is to discover clusters of words or topics, for which many algorithms exist in the literature. Yet most of these methods lack the ...
Extracting time series variation of topic popularity in microblogs
iiWAS2018: Proceedings of the 20th International Conference on Information Integration and Web-based Applications & ServicesExtracting topics and their popularities in microblogs is a promising approach to discover popular topics in the world. To challenge this task, some methods that estimate popularity of topics based on Latent Dirichlet Allocation (LDA) has been proposed. ...
Soft clustering criterion functions for partitional document clustering: a summary of results
CIKM '04: Proceedings of the thirteenth ACM international conference on Information and knowledge managementRecently published studies have shown that partitional clustering algorithms that optimize certain criterion functions, which measure key aspects of inter- and intra-cluster similarity, are very effective in producing hard clustering solutions for ...
Comments