Universal Question Answering in 100+ Languages

Établi : July 1, 2019

Question and Answering (QnA) aims to automatically answer natural language questions posed by human.

Motivated by the goal of providing Universal QnA experience to all Bing users,  we built an universal QnA ranking pipeline from scratch based on a dozens of cross lingual technologies.

With only training data in English language, we manage to ship the QnA feature to 100+ languages and 200+ markets/regions in very short term.

 

Example Cases:

Arabic:

Norwegian:

Dutch:

Russia:

 

Telugu:

Frisian:

Personne

Portrait de Ming Gong (YIMING)

Ming Gong (YIMING)

Principle Applied Scientist Manager, Microsoft STCA Search & Distribution

Portrait de Linjun Shou (寿林钧)

Linjun Shou (寿林钧)

Principal Applied Scientist Manager, Microsoft STCA