Talking to Sibilla: An approach to context dependent natural language comprehension

An important aspect of an artificially intelligent library management system is that it should have natural language comprehension capabilities, so that the user may dialog with the system in his native tongue. By this it is meant that the system, after examining the query, will take the required action, i.e. furnish the bibliographic indications. This would allow the insertion of requests in a helter-skelter fashion, without having to use predetermined frames. The aim of this paper is to present an approach to natural language comprehension based on statistical inference and pattern recognition rather than the more usual linguistic approaches.