This paper presents a framework that uses information fusion to capture similar contexts, and then apply these to learn similar instances from a knowledge base in an unsupervised way. These experiments are part of an initiative to build an intelligent business information system with capabilities for multi-faceted repeatable data analysis and decision making. The proposed framework consists of three components: Query Understanding, Information Fusion and Reasoning & Learning. As part of the proposed framework, we present a new approach to performing the weighting of query terms which is aimed at improving our understanding of a user's query intent. The proposed query terms weighting method captures the key contexts of the user's query intent using evidence from corpus statistics. By way of example, the datasets used in our experiments consist of the information retrieved from different sources pertaining to Mobile Payments, a rapidly evolving sector of the Financial Services industry. We illustrate the performance of the proposed information retrieval system using the new query terms weighting approach on three different datasets. Our experiments illustrate that the proposed query terms weighting approach significantly improves the retrieval of texts with a greater variety of contextual information.