BGU students Neta Weiss and David Moshkovich and their advisors Prof. Mark Last and Dr. Marina Litvak have developed an automatic text summarization website for academic papers they have called SUMMY.
Knowledge. We all need it. But how do we acquire it? Not so long ago, the problem was finding the right information for your needs. The hard task was to find the right articles or academic papers. Nowadays, in the era of Big Data, there is an overwhelming amount of information. In particular, there is an ever increasing number of academic papers and other source material. Now, more than ever, it is of utmost importance to make the process of knowledge extraction more efficient. There is a need to develop a way to efficiently get to the essence of the information, and extract knowledge out of it. To that end, they have developed SUMMY.
The project offers:
Summarization – A sophisticated automated summarization program that has demonstrated outstanding performance in the summarization of articles, academic papers and even books. It is available in 3 languages: English, Hebrew and Arabic. Based on the MUSE algorithm and implemented in C#, it makes the process of knowledge extraction more efficient.
Customization - Different users have different needs. An interface enables the user to personally customize the summarization output.
Website - As part of their vision of accessibility, they have developed a client-server based website wrap-around to the program, making it available to everyone, everywhere.
The brain behind the program is the sophisticated automated text summarization algorithm MUSE. MUSE was developed in the Department of Information Systems Engineering by Prof. Last and Dr. Litvak. The algorithm uses a linear combination of 31 statistical metrics in order to rank the sentences of a given document, extract the sentences which were ranked highest, and generate a relevant summary.
It has demonstrated groundbreaking performance in the summarization of text documents. The original algorithm has been enhanced for summarization of academic articles. Implemented in HTML5, with the design tools of CSS3 and JQuery, the website has a friendly and inviting interface, providing the best UI& UX (User Interface & User Experience).
“To evaluate our objective summarization performance, we have developed a program that takes corpora of academic papers that were summarized by experts, and compares them to the summarization output of our program, in terms of recall and precision,” explains Moshkovich.
“Having a well performing summarization program is great. But it's not enough! We are following our vision of providing the best experience of better and faster knowledge. In order to improve our performance, usability and overall experience, we are currently asking for feedback among BGU students. A cloud-based database is used to collect feedback. Feedback is gathered and
incorporated in order to fine-tune the product. For continuous improvement, we are constantly adding new features while improving performance and usability,” says Weiss.
Check out SUMMY and come see their project in person at Innovation 2014 to be held June 18, 2014 at BGU.