Performance Evaluation of an Improved Model for Keyphrase Extraction in Documents

Awoyelu I.O., Abimbola R.O., Olaniran A.T., Amoo A.O, Mabude C.N.
2016 Computer Science and Information Technology  
Keyphrases are one of the most important parts of a document that give an insight on how a specific document is related. Keyphrase extraction systems are becoming increasingly vital in extracting quality keyphrases. They extract quality phrases that describe a document at hand. Existing keyphrase extraction systems, that employed the unsupervised approach, extract non-domain-specific keyphrases, thereby producing generic keyphrases. An improved model for domain-specific keyphrase extraction in
more » ... ournal articles is therefore proposed in this study. It is a framework that employs document structure, term frequency and inverse document frequency, noun phrase identifier and domain knowledge for keyphrase extraction. Data used in this research include nouns and stop words in English Language. Author-assigned keyphrases were extracted from the International Journal of Data Mining and Knowledge Processing (IJDKP) between the year 2011 and year 2014 for building the domain knowledge and testing the system. It was implemented using Java programming language and MySQL query language. Evaluation was carried out using precision, recall and f-measure as performance metrics. The results obtained show that the proposed system yielded an average precision, recall and f-measure of 27%, 53% and 35% respectively compared to the existing model -MAUIwhich yielded average precision, recall and f-measure of 23%, 45% and 35% respectively. This shows that the proposed model outperformed the existing model by 5%.
doi:10.13189/csit.2016.040106 fatcat:nxdgmv6tufb5ta27lswdlwzs3a