Comparing Different Techniques of Urdu Text Summarization

Automatic summarization is the process of generating condensed version of text. A wide variety of data is available on the web, and to make reading quick and efficient for people, summarization is required. The challenge that comes with summarization is to retain the actual meaning and yet reduce the size of the text. This can be done in two ways, abstractive and extractive. This research focuses on extractive summarization. News is one thing that should be available to be read on the go, and this research works on the summarization of Urdu news. Different extraction based summarization algorithms are used to compare the one that is best suited for Urdu data, given the same data and circumstances. As a part of this research, a sentence segmenter is also created that identifies the ending word of a sentence, since proper sentence boundaries need to be defined to proceed with summarization. The results of all these have been compared with and without preprocessing the input, and the evaluation methodology is equivalent to ROGUE-I. The results show that there are some algorithms that do work slightly better when given non preprocessed data, but the majority of the algorithms give better results on preprocessed input.