Samsung Research, the advanced R&D hub of Samsung Electronics’ SET (end-products) business, has ranked first in the global artificial intelligence (AI) machine reading comprehension competitions twice in a row.
Samsung Research placed first in the MAchine Reading COmprehension (MS MARCO) Competition held by Microsoft (MS) last month. The same month, Samsung Research also showed the best performance in TriviaQA held by University of Washington, proving the excellence of its AI algorithm.
With intense competition in developing AI technologies globally, machine reading comprehension competitions such as MS MARCO are booming around the world. MS MARCO and TriviaQA are among the actively researched and used machine reading comprehension competitions along with SQuAD of Stanford University and NarrativeQA of DeepMind. Distinguished universities around the world and global AI firms including Samsung are competing in these challenges.
Machine reading comprehension is where an AI algorithm is tasked with analysing data and finding an optimum answer to a query on its own accord. For MS MARCO and TriviaQA, AI algorithms are tested in their capabilities of processing natural language in human Q&As and also providing written text in various types of document. This requires more advanced technical capabilities than SQuAD, which is just answering a simple question after reading a short paragraph in Wikipedia.
For example in MS MARCO, ten web documents are presented for a certain query to let an AI algorithm create an optimum answer. Queries are randomly selected from a million queries from Bing (MS search engine) users. Answers are evaluated statistically by estimating how close they are with human answers. This is a test designed to apply an AI algorithm to solve real-world problems.
Samsung Research took part in the competitions with ConZNet, an AI algorithm developed by the company’s AI Centre. ConZNet has skilful capabilities by adopting the Reinforcement Learning technique, the most advanced Machine Learning AI algorithm. Reinforcement Learning advances machine intelligence by giving reasonable feedback for outcomes as a stick and a carrot strategy does in a learning process. Cutting-edge AI technologies including AlphaGo are upgrading machine intelligence by applying the Reinforcement Learning technique.
With an acceleration in global competition in developing AI technologies recently, contests are widespread in areas of computer vision (technologies to analyse characters and images) and visual Q&A to solve problems using recognised images of characters as well as machine reading comprehension. The Beijing branch of Samsung Research won the International Conference on Document Analysis and Recognition (ICDAR) hosted by the International Association of Pattern Recognition (IAPR) in March, putting them in a top-tier group for global computer vision tests. The ICDAR is the most influential competition in Optical Character Recognition (OCR) technologies.
Head of Language Understanding Lab Jihie Kim at Samsung Research says that the question of how AI technologies understand human dialogue and queries to suggest an optimum answer is one of the hot topics in the AI industry. He explains that the Language Understanding Lab at Samsung Research AI Centre is also striving to develop the technology behind an AI algorithm that can talk with people naturally and propose solutions to a problem. He says, “We are developing an AI algorithm to provide answers to user queries in a simpler and more convenient way in real life. Active discussion is underway in Samsung to adopt the ConZNet AI algorithm for products, services, customer response and technological development.”
MS MARCO and TriviaQA are among the top five global competitions in machine reading comprehension. AI algorithms are tested on whether they can understand and analyse questions to offer answers. Those tests are designed by referring to internet users’ queries and search results. The ConZNet algorithm developed by the Language Understanding Lab at Samsung Research is upgrading its intelligence by considering real user environments. The algorithm takes natural language into account such as how people deliver queries and answers online. We could win those competitions because MS MARCO and TriviaQA competitions are about AI capabilities in real user environments. In truth, our algorithm was a bit behind other competitors in tests requiring a simple answer to a question after analysing a short paragraph. But because such technologies have low relevance to real environments of using AI technologies, we are focusing on tests proceeding continuous R&D.
An Open Lab event was held recently to introduce the labs at Samsung Research to other departments in Samsung Electronics. At the event, in-depth discussions were had with engineers in the home appliances and smartphone departments about AI algorithms. Departments dealing with customer services also showed interest because AI-based customer services including chatbots are emerging as a hot topic.
Justin Hume, Chief Marketing Officer for Samsung South Africa, says, “The strengths of Samsung in the AI industry is that we can build a knowledge system about connections between machines and applications and customer demands in the internet of things (IoT) environment comprised of personal devices, based on Samsung Electronics’ diverse product line-up. This will help us to achieve the goal of realising a user-oriented AI system by collaborating with global partners in the industry. Samsung Electronics recently began to launch global AI Centres and we will lead the efforts of working with AI experts at the new centres abroad.”
ConZNet is an acronym of “Context Zoom-in Network”. The name implies that understanding the context of what people say is critical. The advancement of AI technologies will help to understand and analyse short sentences. AI algorithms also need to have capabilities to analyse real-time news reports rather than existing data to give answers to customer queries. Samsung is also developing technologies that an AI algorithm can answer “There are no proper answers to your query.” as well as searching for right answers. The so-called “rejection problem” is an AI technology with high level of technical difficulties.
