Ping An Group
05 Dec 2018

Ping An GammaLab Wins Global AI Machine Reading Comprehension Competition

Best-in-class machine reading comprehension technology set to improve cost-efficiency and enhance customer experience in insurance and banking industries

(Hong Kong, Shanghai, 5 December 2018) Ping An Insurance (Group) Company of China, Ltd. (hereafter “Ping An” or the “Group”, HKEX: 2318; SSE: 601318) is pleased to announce that OneConnect, a subsidiary of the Group, ranked first in one of the world’s most authoritative machine-reading comprehension challenges — the Stanford Question Answering Dataset 2.0 (SQuAD). GammaLab Institute of Artificial Intelligence (GammaLab), owned by OneConnect, scored 83.435, close to the human performance level of 86.831, way ahead of other companies in the challenge. 

SQuAD2.0 is comprised of more than 100,000 questions with over 50,000 new, unanswerable questions written by crowdworkers to look similar to answerable ones. To do well on SQuAD2.0, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering.

According to Ping An, its AI model could be used in various scenarios to respond to customer inquiries.

For example, when agents leave a job, their clients are looked after by newly appointed agents. The handover process takes time as the new agents need to study different back-dated policies. GammaLab’s model will greatly enhance the reading comprehension for agents to help improve their efficiency in customer service.

The model could also be used in the banking insurance scenario, as it takes time to train banking brokers on the different types of insurance products. GammaLab’s reading comprehension skills could save the brokers’ time in providing qualified services for potential customers. In July, OneConnect officially launched intelligent marketing solution Gamma eExpert based on its advanced voice recognition technology and reading comprehension technology, helping sales to shorten their service time by 30% in average and reducing the disputes caused by human errors.

Another scenario is internet arbitration in universal financial inclusion. Small loan companies tend to turn to online arbitration, which is expensive and takes time to resolve, under the current peer-to-peer lending market. With the reading comprehension skill of GammaLab, the arbitrator will finish a case quicker, reducing the cost for arbitration.

Jessica Tan, Deputy CEO, COO and CIO of Ping An Group, said, Our consistent focus and investment in technology R&D is paying off. Ping An will fully utilize its proprietary AI technology in several core areas of the banking and financial sector, further paving the way for an increasingly broad application of the technology.

This is not the first time GammaLab has finished ahead of its rivals. Established less than two years ago, GammaLab has already achieved great results. In May 2018, GammaLab's micro-expression recognition technology made a major breakthrough, topping the list on both Arousal(emotional intensity) and Valence(positive and negative) in emotional intensity, ranking first globally in the One-Minute Gradual Emotion Challenge by the University of Hamburg. In June 2018, GammaLab also ranked first with an accuracy of 94.46% in the EmotioNet Challenge by the Ohio State University.

Ping An is committed to becoming a world-leading technology-powered financial services group. Striving to capture the opportunities arising from the development of smart technologies, the Group will further its strategy to pursue “finance + technology” and explore “finance + ecosystem” to achieve technology-powered business transformation, optimize products and service experiences, maintain stable and healthy business growth and create greater value for its customers and shareholders.


This website uses cookies to help us provide you with better experience and allow us to improve our service. By continuing to browse the site, you understand and agree to our Privacy Policy and Terms of Use .

This website is not supported by IE. Get the latest version of Firefox or Chrome for better browsing experience.