Document summarization by sentence extraction using a genetic algorithm

Text Summarization is a research domain that attracts many research groups around the scientific world. It is the process of automatically creating a condensed version of a given text that provides useful information for the user. Semitic language processing in general is of great interest today. However, the Hebrew language has been relatively little studied. In this research, the application domain is articles referring to Jewish law written in Hebrew. Summarization of these documents is done by extraction of the most relevant sentences. We have developed seven general baseline extraction methods and two specific Hebrew methods. Using a genetic algorithm, these methods are combined into a hybrid method. The success rate of the GA is reasonable compared to the rate achieved by other summarization systems (although they involve different languages, domains and features). This model in general and the baseline methods in particular can be extended with a reasonable effort for documents related to similar domains written in other languages. Investigating other extraction methods and other ML methods might lead to improved.