Semi-supervised learning framework for knowledge extraction in Cricket domain

Abstract

Today’s web is overwhelmed by the data and it is continuously growing, therefore processes of information retrieval and analysis have become tedious tasks. Although many machine learning approaches have been applied to mine these data, many of them are hardly succeeded in their approaches because they have restricted themselves into targeted or downloaded databases. Growing web can not simply be classified or mined by using static knowledge base, system has to grow with the web. Therefore, a system that can mine while learning from the mined data, is required. This paper proposes a framework that acts as a learning model to derive information by building relationships between different entities in online content by relying on few seeds being fed to the system at the start. Couple of extractors are used to derive facts based on their mutual correlations. Those facts have been occupied an ontology to generate new relationships and entities as candidates. A query system has been embedded to the miner to enable querying the knowledge base to retrieve appropriate outputs corresponding to a particular query. The system has evaluated against the cricket online sources.