Name:GNALIS
Description:Grid technologies for natural language interaction systems
Abstract:The aim of the experiment is the exploitation of computational grid technologies for the implementation of natural language interaction systems. The grid will be used to reduce the computational complexity required by methodologies of semantic analysis and natural language representation. The aim is also to obtain a better organization and retrieval of informative contents requested by users.<BR/><BR/><BR/><BR/>The use of natural language in a human-computer interface allows a fulfilling interaction and a greater accessibility to the system by expert and inexpert users. For these reasons in last years there has been a growing interest toward the use of Chatbots[1]which are simple conversational agents using pattern matching rules to carry on a dialogue with the user. <BR/><BR/><BR/>The proposed application is a framework for a chatbot-based information supplier which tries to overcome the rigidness of chatbots dialogue mechanism adding some sort of intuitive, associative reasoning ability. This ability is obtained representing the chatbot KB in a high dimensional vector semantic space built by means of Latent Semantic Analysis(LSA) methodology[1]. The use of LSA allows to infer latent relations between natural language elements belonging to the chatbot Knowledge Base(KB). User questions are mapped in the same semantic space and compared during the dialogue with the chatbot KB documents, in order to obtain the most appropriate answer, by means of a suitable vector similarity measure.<BR/><BR/><BR/>The effectiveness of the proposed framework is influenced by the size of chatbot KB, which has to be wide enough to guarantee a dialogue with the user. However the computational complexity requested for the semantic space building grows considerably with the increase of the analyzed documents. A parallel implementation of semantic space creation process where representing the chatbot KB, allows to obtain a distribution of the working load on a cluster of processors and therefore a reduction of the computational complexity. <BR/><BR/><BR/>Moreover

Created:2010-05-01
Last updated:2012-05-31