Answering complex questions in natural language using probabilistic logic programming and the web

We present an algorithm to answer complex questions in Natural Language that are a boolean combination of simple questions. The main feature of this algorithm is to use a probabilistic Prolog (ProbLog) to handle the uncertainty of answers obtained by information extraction systems such as TextRunner. The results shown in this paper indicate that the use of ProbLog improves the answers given by a system that parses complex questions with light linguistic mechanisms. Therefore, using a probabilistic setting can be seen as a promising approach for the enhancement of open information extraction systems based on weak linguistic algorithms.Oramas Mogrovejo J.A., De Raedt L., ''Answering complex questions in natural language using probabilistic logic programming and the web'', Online Proceedings 22nd Benelux conference on artificial intelligence - BNAIC 2010, October 25-26, 2010, Luxembourg. http://bnaic2010.uni.lu/proceedings.htmlstatus: publishe