Automatic Keyword Extraction From Dialogue Text

Keyword extraction is the automatic process of extracting keywords or keyphrases that are most relevant to a text using various algorithms. In this study, we have extracted keywords from customer service records of various companies. We reached %70 recall score by supporting tf-idf with textrank and positionrank algorithms. This corresponds to %10 relative improvement to tf-idf alone. This method is created to get accurate results especially from small-medium length and preferably dialogue texts. The system can be applied to different types of texts by optimizing its parameters. Our method also generates minimum three word phrases which is an easily understandable and a very short summary of the dialog, from the extracted keywords.