Boston text analysis Conference Report
….Men on a Mission
Intrepid Researchers Hardeep Johal and Nigel Shanahan visited Boston (USA) recently to examine the marketplace for text analysis – the automatic extraction of meaning and sentiment from words. The guys were fortunate enough to witness the Boston Celtics NBA World Series Win over LA Lakers at the same time. They also inform me that Boston Lobster is ‘excellent’, and the Samuel Adams (beer!), ‘superb’!
At some point, they managed to attend a Conference, and to be fair, their findings are very interesting. text analysis is the most exciting thing to happen in the field of market research/analytics and indeed, ‘Voice Of Customer’ - period. Prior to text analysis the focus has been on the 20% of data that was ‘structured’, now it appears we have the tools to work with the other 80% of the data which is far richer in many ways.
Rant & Rave has a Sentiment Engine, and over the next few months, many new enhancements and features are planned. Imagine the power of receiving email and SMS feedback from Customers and being able to automatically derive meaning and Sentiment – a Dashboard with a score between 1-5, and categories automatically showing where things are going well, and where more work is required.
The increasing importance of analyzing social media – consumer blogging, social networks, community boards, and wikis – is changing how firms measure marketing effectiveness. Catching the Voice of the Customer through Surveys and Web 2.0 data through sentiment analysis, and coupling that with existing structured data, such as demographics and transactional history, permits more accurate results, better predictive modelling and customer understanding.
The essential purpose of the Sentiment Engine is to score and categorise unstructured textual feedback from customers. To do this the engine currently has two modules. The first module calculates a score out of 5 for the text received. This is uses Natural Language Processing techniques to spot patterns in the text. The second module parses the text to detect a category to assign to the item of feedback. This is performed by searching the text for keywords that are linked to categories. For example ‘staff’, ‘employee’ and ‘clerk’ can all be used to define the category Staff – the categories and keywords are manually defined by the Rant and Rave customer.
We are also making advances with regard to collecting and evaluating content from the world wide web – the main purpose of our visit to the Conference, actually. For a given topic we search for sites that commonly contain content regarding that topic and subscribe to the RSS feeds that belong to the sites. The content is stored in a database ready to be passed through the Sentiment Engine.
If you would like a stimulating half hour with Hardeep on Word Sense Disambiguation, Lexicons, Taxonomies and Named Entity Extraction, then please do not hesitate to get in touch!
For more information on Rant & Rave’s Sentiment Engine please provide us with your name, contact number and email address and we will contact you shortly.
Click here to forward this article to a friend >>
Click here to return to the front page >>
