Semantic Representations for Textual Inference
     Natural Language Technology is moving from text retrieval and search applications to tasks that require genuine understanding of natural language as well as the interaction between language understanding and reasoning. Within the linguistic computational world a common perspective has emerged on what is common to these natural language understanding tasks under the heading "textual inferencing". The aim is to develop systems that can decide when given two natural language statements, what the inferential relation between the two is. Textual inference simplifies the general language understanding problem by limiting its interest to direct inferences avoiding complicated chains of inferences and specialized world knowledge. But even within this narrow conception of inferencing it remains to be seen how we can get an inferential relation between two texts (e.g. a question and an answer). Several approaches have tried to avoid deep linguistic analysis altogether but others rely on a semantic representation built on top of a syntactic parse. Within the latter type of approaches there are alternatives both in the actual form that the semantic representation takes and in the way it is arrived at. The current workshop aims to present and discuss some of these alternatives.

This March 9 -10 , 2012 event is sponsored by
the National Science Foundation
the Center for the Study of Language and Information
and organized by the "Language and Natural Reasoning" group at CSLI.

Earlier recent workshops on similar topics and organized by members of the same research group can be found here and here .