Topic Assisted Fusion to Re-rank Texts for Multi-faceted Information Retrieval

Abstract

We propose to develop a framework for an intelligent business information system with multi-faceted data analysis capabilities that supports complex decision making processes. Reasoning and Learning of contextual factors from texts of financial services data are core aspects of the proposed framework. As part of the proposed framework, we present an approach for the ordering of contextual information from textual data with the help of latent topics identified from the web corpus. The web corpus is prepared by specifically using a number of financial services sources on the web that describe various aspects of mobile payments and services. The proposed approach first performs weighting of query terms and retrieves the initial set of texts from the web corpus. We use Latent Dirichlet Allocation (LDA) on this web corpus to identify the topics that relate to the contextual features of various financial services/products. The retrieved texts are scored based on the identified topics that could cover a variety of contextual factors. We performed subjective evaluation to identify the relevance of the contextual information retrieved, and found that the proposed approach captures a variety of key contexts pertaining to user information needs in a better way with the support of topic assisted contextual factors.