Name Entity Detection and Relation Extraction from Unstructured Data by N-gram Features on Hidden Markov Model and Kernel Approach

Abstract

In recent years Name entity extraction and linking have received much attention. However, correct classification of entities and proper linking among these entities is a major challenge for researcher. We propose an approach for entities and their relation extraction with feature including lexicon, n-gram and parts of speech clustering and then apply hidden markov model for entity extraction and CRF with kernel approach to detect relationship among these entities. Analysis of our model is done by precision, recall and accuracy. We have used kernel approach with Conditional random field for extracting the relation between the entities and then remove the co-reference by kernel function. The accuracy of the proposed system for entity detection is 98.03, precision is 88.80and recall is 87.50 where as accuracy of relation extraction is 87.46,precision 84.46 and recall is 82.46 which is much better than the rest existing models.

4 Figures and Tables

Cite this paper

@inproceedings{Priya2015NameED, title={Name Entity Detection and Relation Extraction from Unstructured Data by N-gram Features on Hidden Markov Model and Kernel Approach}, author={N . Swathi Priya and Amanpreet Kaur}, year={2015} }