Online social networks has witnessed an unprecedented growth in recent years. At the same time the social network services are exponentially pervading through the various sections of our society, the value of the information produced by such systems dramatically increases. Indeed, interactions and communications in online social networks often reflect real-world events. Among them, Twitter is the most popular microblogging service with nearly 600 million users and produces more than 300 million tweets per day. In addition to information sharing, users also refer to Twitter for information about news and real-time events, as Twitter search engine is accounted for more than 2.1 billion queries per day.

The high speed and the volume of production and dissemination of information in these networks has led to an "information overload" phenomena that makes processing the information difficult if not impossible for users. Thus, automatic methods are required for organizing and summarizing the timeline of news events to facilitate users' access to the relevant information.

Our goal in this research is providing a comprehensive picture of the most important events of a news topic, as well as the relationship between them. This big picture, allows users to browse news topics and related events and also manifests the hidden relationships among them, especially in complex news topics.

In this research, the problem of timeline summary generation of news events in microblogs is explored. Although, few attempts are performed at producing structured summary of news events in Twitter, each of them has limitations and shortcomings. In this study, we sought to overcome these challenges.

Toward this, our approach consists of a 3-step process to solve timeline summary generation of news events in microblogs. First, relevant tweets to user query are retrieved utilizing query expansion techniques to improve the accuracy of retrieval. Second, the problem of timeline generation on retrieved tweets is formulated through an optimization framework. Finally, a summary of tweets related to each step is generated.

Provider

نازنین دهقانی email: nazids [at] gmail.com

 

Supervisor

مسعود اسدپور email: asadpour [AT] ut.ac.ir