AITRICS to Publish a Paper in NAACL, International Society for Natural Language Processing
[May 23, 2022] AITRICS (CEO Gwangjoon Kim, Jinkyu Yoo), a company specializing in artificial intelligence (AI) technology, held the International Natural Language Processing (NLP) conference 'NAACL 2022 : Annual Conference of the North American Chapter of the Association for Computational Linguistics' and announced on the 23rd that it would present a paper on a newly developed language model. NAACL is considered one the best international academic conferences in the field of natural language processing along with with Association for Computational Linguistics (ACL) and Empirical Methods in Natural Language Processing (EMNLP). AITRICS will participate in the NAACL 2022 in Seattle, Washington, from July 10 to July 15 and present a paper titled 'KALA: Knowledge-Agumented Language Model Adaptation'. The KALA framework, presented by AITRICS in this paper, showed higher performance than the existing language model in various fields of natural language processing tasks, including question-and-answer in Electronic Medical Record (EMR) required in the medical field and disease entity recognition in medical papers. Existing pre-trained language models (PLMs) such as BERT perform well on various tasks such as machine reading and object name recognition, but natural language processing performance needed to be further improved in special areas such as healthcare. Several methodologies were presented to solve this problem, but there was a disadvantage that it still required a lot of calculation costs.The KALA framework proposed by AITRICS in this study overcomes this limitation. The KALA framework has the advantage of utilizing knowledge graphs to improve the performance of language models in special fields at a low computational cost. Based on these advantages. Based on these advantages, AITRICS's paper was also invited to an Oral Presentation session, where only a few outstanding studies were given opportunities among the accepted papers. The study was jointly researched by Min-ki Kang, a researcher at AITRICS, Jin-heon Baek, a Ph.D. course at KAIST, Sung-joo Hwang, general manager of AI Division at AITRICS, and professor at KAIST AI Graduate School. Min-ki Kang, a researcher at AITRICS said, "The KALA framework developed by AITRICS can improve the computational efficiency of adapting language models to special fields by mapping entities in unlearned fields to embedding spaces", and In the future, AITRICS expects to solve unmet medical demand in the medical field through various research and development in a wide range of fields, including natural language processing". Meanwhile, AITRICS is a startup specializing in artificial intelligence(AI) technology composed of the best machine learning R&D personnel, and is recognized worldwide for its remarkable excellent performance every year by various international AI societies such as NeurIPS and ICML.