Pre-symposium tutorial
Practical Sentiment Analysis
New in 2011, the Sentiment Analysis Symposium will be preceded Monday afternoon, April 11 by a three-hour tutorial, Practical Sentiment Analysis, designed for and taught by sentiment-analysis practitioners. You may register for the tutorial or the symposium or both.
Instructors
Yongzheng Zhang, Catherine Baudin, and Nitin Indurkhya, eBay Research Labs
Syllabus
Sentiment Analysis, also known as Opinion Mining, is the computational study of sentiments and opinions expressed in unstructured text documents. This field has been rapidly growing in the last decade. Blogs, social media, editorials, product reviews, and even traditional newspaper articles contain many opinions and sentiments. These relate to a questions of great interest to service providers and product manufacturers. For example, an e-commerce site may want to ask: how do users like the new search function? Or a phone manufacturer might want to know: how do people like their latest smartphone? More diverse questions can be answered from relevant text documents: How does the community react to new procedures to construction permit applications? Is there a change in the support rate for the new medicare system? In the product review area, end users will benefit greatly from sentiments in product reviews while conducting shopping research and manufacturers will find the sentiments helpful for marketing and product improvements. The source texts contain a lot of information that is not directly relevant to sentiments/opinions. Thus, it is critical to accurately extract and interpret such sentiments from the full text.
In this tutorial, we will focus on sentiment analysis from user-generated content on the Web, e.g. forum posts, blogs, and consumer reviews. We will cover essential components in a sentiment analysis system, including but not limited to the following:
- Topic Extraction: information retrieval and text mining techniques that are used to extract essential topics and sentiment-related vocabulary from a document collection;
- Sentiment Extraction: start-of-the-art techniques that are used to identify opinions and sentiments in a particular topic;
- Opinion Summarization: summarization techniques that can be used to consolidate and summarize multiple sentiments for a topic;
- Opinion Spam: Techniques for isolating spam and untrustworthy opinions in reviews.
While the tutorial is aimed at practitioners, it will also survey promising material that is still incubating in academia. We will illustrate some state-of-the-art academic and commercial sentiment analysis systems and tools. Finally, we will showcase two demos developed at eBay Research Labs addressing sentiment analysis from forum posts and product reviews. The tutorial will consist of 2 sessions of 90 minutes each.
Registration
You may register for the tutorial or the symposium or both. Visit the registration page.