Delta recently placed 5th and won the 3rd prize among 153 international teams in the "2018 NLP Challenge on Machine Reading Comprehension" organized by Baidu, the Chinese Information Processing Society of China, and the China Computer Federation.

During the award ceremony, Team Leader Yi-Ying Tseng received the prize on stage, and represented the 3rd prize winning team to speak on stage. Tseng explained the NLP technologies and machine reading comprehension models to hundreds of industry and academia experts.

Delta team members (left to right): DRC Yi-Ying Tseng, Chi-Chieh Shao, Yu-Ting Lai, and DMS Po-Chen Lin, Rui-Chen Hsiao

Delta employee Yi-Ying Tseng explains the data and model used in the competition to hundreds of industry and academia experts in the audience

The day-to-day research field of Delta’s team, consisting of staff in the DRC Intelligence in Text Team and DMS, includes NLP and machine reading comprehension. In this NLP Challenge on Machine Reading Comprehension competition, the whole dataset came from real questions on Baidu search engine and "Baidu Knows". There were over 270,000 pieces of data, the largest Chinese dataset in history.

After this competition, the team not only received a valuable real dataset, but also experience in data processing and model optimization. "After the external resources and experience we acquired through the competition are combined with the Chinese dataset independently developed by DRC, Delta's machine reading comprehension was reinforced, particularly in the search for professional knowledge in specific fields." says Team member Chi-Chieh Shao.

The next step for the team is to focus on a machine reading comprehension model that is knowledgeable and can understand questions made by real people. Along with the word generation technology, users are allowed to obtain answers quickly from the database in a user-friendly and intuitive Q&A window. It could further reinforce Delta's intelligence and knowledge management capabilities.