A comparison of feature and semantic-based summarization algorithms for Turkish

Abstract

In this paper we analyze the performances of a feature-based and two semantic-based text summarization algorithms on a new Turkish corpus. The feature-based algorithm uses the statistical analysis of paragraphs, sentences, words and formal clues found in documents, whereas the two semanticbased algorithms employ Latent Semantic Analysis (LSA) approach which enables the selection of the most important sentences in a semantic way. Performance evaluation is conducted by comparing automatically generated summaries with manual summaries generated by a human summarizer. This is the first study that applies LSA based algorithms to Turkish text summarization and its results are promising.