Design of an efficient information retrieval system for query retrieval

Abstract

Exponential growth in digital data results in vast amount of information. The availability of large volume of information results in information overload leading to increase in time and cognitive resources to retrieve the most relevant information. Enterprise information retrieval systems still use specific keyword base search which fail most of the time in retrieving the most relevant information in lesser time. Enterprise retrieval is a broad area and user expectations are quite high in enterprise retrieval. newline newlineEnterprise information retrieval is quite challenging than web-based information retrieval due to retrieval from heterogeneous distributed data sources and need for higher accuracy in retrieval. Large corpus of data from heterogeneous information sources available as distributed systems make querying for most relevant and updated information in less number of trials complex. newline newlineThis research applies collaborative fusion-based information retrieval applying three dimensions of user similarity, user-document search history and document similarity. Concepts were extracted from document and user query is mapped to concepts. Documents similar to search query concepts are found applicable to the three dimensions as discussed earlier work. Improvising the search results based on feedback is also proposed to re-rank search results for a personalized view. The proposed retrieval is extended for fine grained access control required in enterprise environment, with zero leakage assurance through direct or inference. In order to make the retrieval process work efficiently even in the presence of cold start problems, user profile modeling from enterprise communication systems is also proposed in this research work. newline newlineThe proposed solution is found to have a precision of 91% an increment of 6% compared to 85% of earlier solution. The proposed solution has a ranking accuracy of 83% compared to 73% in earlier solution. newline newline