Named Entity Classification Using Search Engine's Query Suggestions

Jayendra Barua, Dhaval Patel

Abstract: Existing Named Entity Recognition (NER) techniques uses external gazetteers lookup as a feature to improve classification accuracy of entity mentions in the text. However, gazetteers lookup suffer with low recall problem as creation and maintenance of gazetteers is a labor and cost intensive task. In this paper, we propose to use Search Engine’s Query suggestion as external knowledge source instead of gazetteers for named entity classification in NER systems. Specifically, we build a Query Suggestion based Named Entity Classifier (QS-NEC), which learns entity types from Query Suggestions of Named Entities. We have used QS-NEC as an Entity Classification module in our NER framework. Our experiments on MSM Challenge dataset demonstrate that QS-NEC is efficient in classification of entity mentions and can be effectively used in NER systems.