The first was the MAchine Reading COmprehension Competition (MARCO) arranged for by Microsoft. Samsung’s second similar achievement via displaying excellence in AI algorithms came in TriviaQA hosted by the University of Washington.
These two competitions sit with NarrativeQA of DeepMind and SQuAD of Stanford University to be the actively researched and used machine reading comprehension contests. Global AI firms and distinguished universities around the world are participants.
Machine reading comprehension is most simply defined where an AI algorithm is tasked with “analyzing data and finding an optimum answer to a query on its own accord.” In the two contests where Samsung prevailed, the algorithms are tested on capabilities of processing natural language as humans in questions and answers. The tests also involve providing written texts in different document types like blog posts and news articles.
In MS MARCO, ten web documents were presented to Samsung’s AI algorithm for creating an optimum answer to a specific query. These queries were randomly picked up from the users of Microsoft’s search engine Bing.
Answers in MS MARCO are statistically evaluated by comparing the counterpart answers from humans. Interestingly, this test has been designed for applying AI algorithm to solve the problems of the real world.
The model of Samsung Research is named ConZNet that’s developed in the tech leader’s AI center. It has the ability to adopt a Reinforcement Learning technique, which is designed to advance machine learning intelligence in providing practical feedback for outcomes.
Samsung has another crown in the same field, and that has come from it winning the ICDAR (International Conference on Document Analysis and Recognition). This was hosted by IAPR (International Association of Pattern Recognition) in March this year. It is a top-tier global computer vision test and is the most influential contest in OCR (Optical Character Recognition) technologies.