One of the internation offices of Samsung located in Sunnyvale, California. / Photo by: Christopher Ulrich via Wikimedia Commons |
These are the Microsoft Machine Reading Comprehension (MS MARCO) competition, and the TriviaQA hosted by the University of Washington, according to Paul Hill, reporting for Neowin.
The Korean technology giant put a lot of significance on the victory of its AI algorithm in the said events.
The world’s leading organizations and universities are using the two tournaments in testing the capabilities of their AI algorithms in processing natural language in human conversations, and for analyzing text in different types of documents
During the MS Marco competition, the ConZNet algorithm was given 10 documents together with a question, and the algorithm must be able to come up with the best answer possible based on the information that it was given.
There was no opportunity to program answers into the ConZNet algorithm because the questions were randomly chosen by Bing users.
To rank the scores of the participants, their answers were compared with those made by humans.
The ConZNet algorithm incorporates skillful capabilities that it had learned using the Reinforcement Learning technique, which provides feedback for every outcome.
The technique was also used in Google’s DeepMind AlphaGo algorithm.
The feat comes in the heels of the success achieved by the Beijing branch of Samsung Research, which emerged victorious during the International Conference on Document Analysis and Recognition (ICDAR) event hosted by the International Association of Pattern Recognition (IAPR) in March 2018.
The ICDAR is the world’s leading competition in computer vision and optical character recognition technologies.
Samsung said the ConZNet algorithm was designed to answer user questions in a simpler and more convenient fashion.
It added that there are plans to integrate the algorithm with Samsung’s product lines, commercial services, customer service and product development.