| 简介： | Abstract: In this talk, I will share some personal experience of utilizing statistical modelling and analysis technique to help in improving accuracy of analyzing large-scale text data. Three examples will be discussed in details, including sentiment analysis, learning to rank, recommender system, among others. Text data often come in the form of texts and can be vectorized into numeric vectors, yet rich structure information is often neglected in this process and the analysis afterwards. We will illustrate that these structure information can substantially benefit the text data analysis when they are appropriately incorporated in the statistical models. Also the size of text data is large-scale due to the rapid advance of information technology, and thus the computational challenges will be discussed as well.
|