Content Mining in Practice for Microblog Search

 

Toru Shimizu, Search and Media Science, Yahoo! Japan

 

Currently, way more than one billion people worldwide are using microblog services such as Twitter, Facebook and Sina Weibo, producing an enormous amount of updates constantly.  They often reflect what users are going through at given moments.  They often reflect users' sentiment over diverse topics.  In these circumstances, providing effective vertical search services fully exploiting those real-time, rich streams of user-generated content is only becoming important.

 

In this talk, I will present the following two novel approaches in microblog content mining for Yahoo! Japan's “Real-Time Search”, which is one of the most popular microblog search services in Japan.

* Extracting trending topics and making them structured and rich

* Incorporating push notifications into search

 

Also, some researches have recently shown intriguing results using recurrent neural networks with LSTM/GRU units for tasks like machine translation or generating sentences from images.  I will briefly discuss the method's applicability to content analysis, based on preliminary results of experiments.