SAMA: A TWITTER BASED WEB SEARCH ENGINE

摘要

How can a model efficiently identify relevant references in the hundreds of millions of Twitter messages that are posted every day? In this paper, we intend to address this fundamental research question, as well as introduce SAMA, a scalable search model that uses Twitter streams. Real-time topic detection is an important function for all search engines, and extracting topics from Twitter raises new challenges. As a huge temporal data flow, Twitter has many various types of topics, as well as a lot of noise. Current sophisticated search engines with high computational complexity are not designed to handle such large data flows efficiently. Twitter provides many opportunities for people to engage with real-time world events through communication and information sharing, as well as tools for dealing with its data. However, little is understood about the external links available in Twitter content, and this affects topic engagement. As of today, Twitter posts and its external links is very limited using upon traditional search engine despite the fact that content of micro-blogging presented by Twitter is very curious and useful for some queries rather than content of traditional Webs. In this paper, we propose a platform for modeling URL and inverse message frequencies and Twitter external references, which allows us to use a novel self-content detection algorithm for link authorities. Our model can make use of a new source of Web references, and experiments verify the effectiveness of the model in real time topic detection of Twitter social content. In our evaluations, we investigate the impact of different features on retrieval performance, and highlight tweet features that have high precision for both adhoc and diversity tasks: 77% and 78% respectively.