Abstract:
Currently the number of electronic data is increasing than ever before and we can find high frequency of named entities in electronic texts. Named entity relation extraction is the process of
finding the relation between two named entities from input text, which is a foundation of semantic
networks, ontology design and widely used in information retrieval and machine translation as well
as question and answering systems. In this study we develop a hybrid approach by combining a
machine learning approach using Support vector machine (SVM) and set of rules. We first used
the classifier to predict relations found between named entities. And then to improve the result
which is obtained from the machine learning component we used set of rules. Precision, recall
and f-measure are used to measure the performance of our proposed system. We have used a total
of 764 annotated sentences for training and testing purpose. Our testing is conducted for specific
relationship types separately and the highest precision value achieved in this work is 94% for Àì-
®p , the highest recall is also 96% for E Ì- í5t- and the highest f-score is 92% for Àì- ®p .
To measure the overall performance of the system we take the average value and it gives us 80%,
81% and 83% of precision, recall and f-score value respectively.
