Samsung SDS Tops Both Domestic and International AI Machine Reading Comprehension Competitions

Samsung SDS won several domestic and international AI machine reading comprehension competitions, gaining recognition for its technological capabilities.

+ Samsung SDS takes first place in English and Korean AI machine reading comprehension (MRC) competitions,
proving its technological capabilities.
+ World renowned AI conferences and journals, including NeurIPS, publishes Samsung SDS’s papers
+ To share its technology know-how and application cases at the Techtonic 2020, a developer conference

Samsung SDS won several domestic and international AI machine reading comprehension (MRC) competitions, gaining recognition for its technological capabilities.

The AI-based natural language understanding and learning model developed by Samsung SDS’s Technology Research recently won first place in AI MRC competitions for HotpotQA (tests English) and KorQuAD (Korean Question Answering Dataset) 1.0 and 2.0 (tests Korean).

HotpotQA is a dataset-based global AI engine competition developed by Carnegie Mellon University, Stanford University and Université de Montréal. Samsung SDS ranked top in the HotpotQA Fullwiki Setting, the most difficult category with answers found throughout all of Wikipedia. AI engines are asked to find answers based on two or more sentences which requires advanced and logical reasoning.

In the KorQuAD 1.0 competition, AI engines are asked to process a limited amount of information to find accurate answers to given questions. Meanwhile, answers to the KorQuAD 2.0 test are required to be found throughout all of the Korean Wikipedia, and AI engines have to answer in long sentences which requires an advanced level of understanding of sentences. In particular, the Samsung SDS team scored higher than human scores among the KorQuAD 2.0 contestants for the first time.

In addition, 11 papers announced by Samsung SDS teams were published by world renowned conferences and journals, including NeurlPS, ACCV, AAAI, ICIP, and ICPR.* In particular, in September, Samsung SDS’s paper about a reinforcement learning algorithm to optimize product design, manufacturing and logistics processes was published by NeurlPS 2020, one of the most prestigious AI societies in the world.

NeurlPS: Neural Information Processing Systems
ACCV: Asian Conference on Computer Vision
AAAI: Association for the Advancement of Artificial Intelligence
ICIP: IEEE International Conference on Image Processing
ICPR: International Conference on Pattern Recognition

Samsung SDS is focusing on securing the best talent in AI. Recently, Samsung SDS scouted an AI multi-modal (voice, text, video, image) expert from the MIT Lincoln Laboratory and a former Google expert in computer vision as a leader of AI Research, further upgrading the company’s AI technological power.

Samsung SDS has applied such strong AI capabilities to a wide range of areas: manufacturing (process improvement, root cause analysis, improvement in the accuracy of defect classification); retail (sales forecasting, marketing effect analysis); finance (analysis of accounts’ financial solvency, automated trade audit); and healthcare (electrocardiogram analysis and ocular fundus video interpretation).

Samsung SDS will host the 3rd developer conference, Techtonic 2020, online on November 12. The event will feature presentations and discussions about core R&D know-how and application cases regarding key technologies, including AI, Blockchain, cloud, data analysis and security.

Sang-wook Lee, Technology Research Leader and Senior Vice President of Samsung SDS said, “We will advance our AI technology research and continue securing diverse talent to actively support the digital transformation of our customers.”