Combining Information Retrieval with Information Extraction for Efficient Retrieval of Calls for Papers

In many domains there are specific attributes in documents that carry more weight than the general words in the document. This paper proposes the use of information extraction techniques in order to identify these attributes for the domain of calls for papers. The utilisation of attributes into queries imposes new requirements on the retrieval method of conventional information retrieval systems. A new model for estimating the relevance of documents to user requests is also presented. The effectiveness of this model and the benefits of integrating information extraction with information retrieval are shown by comparing our system with a typical information retrieval system. The results show a precision increase of between 45% and 60% of all recall points.