Keyword extraction from abstracts and titles

Keywords are very important for any academic paper. We propose the Perceptron Training Rule for keyword extraction from titles and abstracts. We present a system for generating keywords which relies on weights of words in a sentence. The system generates keywords from academic research articles by selecting the most relevant keywords. We compare the keywords generated by our system and those generated by cluster analysis to the keywords given by the authors and analyze the results based on full-keyword matches, partial-keyword matches and no-keyword matches.