engleski

Automatic Keyphrase Extraction from Croatian Newspaper Articles

Keyphrases provide a way to summarize documents and enable cross-category retrieval. The paper describes a robust system for automatic keyphrase extraction from newspaper articles in Croatian language. Keyphrase candidates are generated based on linguistic and statistical features, and naïve Bayes classifier is used to select the best keyphrases among the candidates. A prediction model is built using training documents with human-assigned keyphrases. System performance is measured on a corpus of newspaper articles, by comparing the automatically extracted keyphrases with those assigned by professional indexers. In absence of comparable results, we consider our results to be of modest performance.

keyphrase extraction; naïve Bayes classifier; Croatian language

nije evidentirano