News Source: School of Computer Science and Technology

Editor: Ling Ting

Translator: News Agency of BIT

Recently, Professor Song Dawei's team from the School of Computer Science and Technology, BIT, has made a series of new progress in large-scale pre-trained language models, including its domain adaptation, knowledge distillation, model compression and downstream task application (such as sentiment analysis, semantic matching, text generation, etc.) On September 25, the team's academic paper was awarded the Best Paper Award of NLPCC2022 (belongs to CCF-C). On October 6, EMNLP 2022, the top conference in the field of natural language processing, released the results. Professor Song Dawei's team performed excellently and a total of four high-quality full papers were accepted.

Sponsored by the Chinese Computer Society, Natural Language Processing and Chinese Computing (NLPCC) is a high-level international conference in the field of Natural Language Processing (NLP) and Chinese Computing (CC). The NLPCC 2022 was held in Guilin from September 22 to 25. The paper entitled "Doge Tickets: Uncovering Domain-general Language Models by Playing Lottery Tickets ", written by Yang Yi and Zhang Chen from professor Song Dawei's team, in collaboration with the Chinese University of Hong Kong (Shenzhen), was selected the best paper.

Conference on Empirical Methods in Natural Language Processing (EMNLP) is the top international conference in the field of natural language processing. Hosted by the Association for Computational Linguistics (ACL), EMNLP ranks second in Google Scholar’s index of computational linguistics publications and was one of the CCF-B recommended conferences (also an A-class conference recommended by Tsinghua University). EMNLP 2022 will be held in Abu Dhabi from December 7 to 11 online and offline. In the meeting, four papers are accepted as full papers, which are written by Professor Song Dawei's students Zhang Chen, Zhang Hanqing, Ma Fang, and Yang Yi.