Agenda

This session will cover how HP is using sentiment analysis to understand what IT professionals are really saying about the company's products. The data that is gathered supports the editorial content efforts of over 150 employees. You will learn the pros and cons of using sentiment analysis, what to look for in a solution and how to promote the adoption and usage of social media throughout an organization.

Sentiment analysis technologies are quite mature today and allows for customization to provide fairly accurate insights about consumer sentiments. However, sentiment analysis is yet to be adopted into main-stream business intelligence since methods to integrate sentiment analysis reports with internal and/or structured data of the organization are not well-defined. In this presentation we share our experience in building methodologies to integrate sentiment analysis results with business data and also discuss a set of key-performance indicators (KPI)s designed that can provide useful insights about the effects of consumer and market sentiments on business. The integration can be done with a large variety of internal data like sales data, market share data, promotion plans etc. Integration can be performed along various dimensions like time, region, product /services, business units, type of information etc. to derive useful insights.

In this presentation, we shall share our experience of a few case-studies in the retail domain. We show how context of consumer comments plays an important role in the analytics process by providing the reasons for sentiments reported. We illustrate through a set of scenarios how consumer sentiments are influenced by external events like product promotions, price rise announcements, product launches etc. We also illustrate how causal analytics is deployed to build predictive models by analyzing the overall impact of past events.

This presentation will review the state of the art in Sentiment Analysis, and highlight some new approaches which represent a quantum leap for the technology.

Many companies offering sentiment analysis are doing little more than matching sentiment-laden keywords. Often the "analysis" of these keywords consists of little more than counting them. These approaches have been shown to be generally too noisy to be useful. A smaller number of companies are taking the challenge of sentiment analysis more seriously, using natural language processing technologies, but almost all of these companies focus on English only or a very limited set of European languages.

We will walk through a number of examples showing the limitations of this kind of approach and highlighting the need for deeper linguistic analysis. We will describe what we would like to call rich sentiment analysis: which combines detailed analysis of sentiment words and their modifiers ("very good", "not good", etc.), as well as highlighting the context of the sentiments - which trending topics correlate with strong sentiments, both positive and negative. We will also discuss the importance of managing the trending topics which organizations want to monitor. This approach shows how Sentiment Analysis technology is a much more powerful tool when focused on strategically relevant areas rather than being applied as blunt instrument.

Furthermore we will show how this deep approach can be applied in multiple languages: English, French, German, Japanese and Chinese.