About WebQA2


  
WebQA2 - Multilingual Web-based Question Answering

© German Research Center for Artificial Intelligence GmbH (DFKI), 2011. All rights reserved.

WebQA2 is a online multilingual web-based question answering engine. It receives natural language questions (not keywords) in one language and returns exact answer strings (not documents) extracted from Web snippets of the same language. WebQA2 answers open-domain factoid Wh-questions of the type Where, When and Who for the languages English (EN), German (DE) and Spanish (ES). WebQA2 answers concept description questions of arbitrary type in English (EN), Spanish (ES), and soon in German (DE). The Web snippets are determined by a external standard search engine. Currently, we are using Microsoft's search engine Bing.

WebQA2 is a research prototype, so the focus is on the investigation and exploration of new methods and algorithm in the area of multilingual web-based question answering. As such, its main goal is to make some contribution for future answer engines.

Our main motivation for the development of WebQA2 is the goal of extracting exact answers for different types of questions from Web snippets only (not from Web pages) using strong Machine Learning based data-driven strategies. We are exploring unsupervised statistical-based methods making use of very lightweight natural language specific data structures only, i.e., stop-word lists and simplistic patterns as main language specific resources.

More information about the underlying technology implemented in WebQA2 and the outcome of detailed experiments can be found in the following publications: