Samsung SDS announced on Oct. 22 its AI natural language understanding learning model recently topped the "HotpotQA," which is conducted in English, and the "KorQuAD: (Korean Question Answering Dataset) 1.0" and " KorQuAD 2.0" AI Reading Contest, which tests the Korean language.
"HotpotQA" is a dataset-based global AI reading contest created by Carnegie Mellon, Stanford and Montreal universities. Samsung SDS topped "HotpotQA Fullwiki Setting," which is the most difficult way to find answers. This test requires AI to provide the correct answer to the question based on two or more sentences.
KorQuAD 1.0 is a test in which AI reads a limited number of contents on suggested questions and finds the correct answer. KorQuAD 2.0 requires long-term answers based on a high level of understanding of sentences in a way that finds answers throughout Wikipedia in Korean. In particular, Samsung SDS was the first team to score above the level of a human being.
Samsung SDS also achieved this year for the adoption of its 11 papers by global AI academic societies such as NeurlPS, ACCV, AAAI, ICIP, and ICPR.
In September, its paper on strengthening learning techniques for optimizing product design, manufacturing, and logistics processes was adopted by 'NeurlPS 2020', one of the AI's most prestigious academic societies.
Lee Sang-wook, director of Samsung SDS' research center, said, "We will actively support customers' digital transformation by upgrading research on AI technologies and continuously securing diverse talents.”