Manya Mayes, a Solutions Architect and Chief Text Mining Strategist at SAS, just gave a wonderful presentation about using SAS Text Miner with SAS Content Categorization. Both products have been used successfully by SAS customers for some time. But Manya and our clients are now using them to solve old problems by using a few new tricks.

Manya invited me to watch the presentation because we're both interested in gleaning as much information as possible from our Twitter and social media conversations. Manya has been doing a great deal of research into sentiment analysis, which is mining data, including messages such as e-mail, text messages, call center traffic and Twitter, to determine how an audience “feels” about a topic. She wanted to show me how her research applies to our shared social media interests and to an organization's need to understand what customers are thinking and feeling.

According to Manya, SAS Text Miner was launched in 2002 with the thought that organizations would use it to mine call center data to find unhappy customers before they terminated their accounts. Mining the call center transcripts for churn reduction made this seem like a great customer relationship management (CRM) tool. However, automotive organizations soon noticed that SAS Text Miner had broader capabilities than searching documents for pre-determined words and phrases. These customers needed to spot trending topics and rank them by importance. They would use the tool to find product malfunctions and excessive wear issues.

SAS Text Miner discovers and extracts information from text documents and then transforms the information into a usable format to allow you to classify documents, find relationships or associations between documents, cluster documents into categories and predict behaviors. In the beginning, organizations wrote the search rules and asked SAS Text Miner to search and classify the data. Later, SAS customers that wanted to discover and rank trends relied on SAS to mine all of the data. The data can arrive from a variety of channels, including e-mails, text messages, Twitter streams and faxes. It can be retrieved in various formats (e.g., PDF, ASCII, HTML, Microsoft Word and WordPerfect) and in numerous languages. Combining structured data and unstructured data allows you to incorporate valuable demographic and behavioral data.

SAS Content Categorization, powered by Teragram technology, takes the entire process one step further: When used with SAS Text Miner, data is categorized and then automatically tagged to provide access and “findability” for faster, more efficient information organization. The Text Miner automatic discoveries augment the Content Categorization linguistic definitions. All I remember of statistics from grad school is manual entry text mining – this removes all of that or gives you the reins when you want it.

In the past year, SAS has released developments that take this idea even further. In March 2009, SAS released SAS Social Network Analysis (SNA), a fraud solution that helps institutions detect and prevent fraud. The solution lets financial institutions go beyond transaction and customer views to analyze all related activities and relationships within a network, including shared addresses, telephone numbers, employment, account ownership and other key transactional data. Marketers in other industries are using SNA to identify customers who have the most influence among their peers. And in June, Teragram released Sentiment Analysis Manager, a social media analysis tool that enables brand managers to see what their end-users are writing about their products in the online world. The sentiment analysis manager is the industry's first system that combines a statistical method for computing reviews with a rules-based approach that lets brand managers evaluate certain specific terms and syntaxes while also allowing for human interaction. When combined with sentiment analysis, patterns found through social network analysis can provide deep insight into both what people think and how they behave, offering predictions not only for what they might want in a product, but what might influence them to buy.

Today’s presentation took us through an unnamed call center transcript. In this case, the callers were complaining about a newly improved product. But what if your organization’s marketing department routinely scanned Twitter searches, forums, blogs, customer response forms and call center transcripts for suggestions or product satisfaction? What would you change about your product line or tech support? How would that change your cross-sales efforts and profitability?

Manya Mayes will present a Text Analytics Lab on Wednesday, Feb. 17 at Predictive Analytics World in San Francisco. The practice run that she presented to the at-home crowd was great, but I'm sure she kept some extra sparkle for those of you who'll attend the event! There's still time to register and SAVE 15% when you use the SAS discount code SASPAW010.

Keep an eye out for SAS Text Miner 4.2! The projected official launch is Feb. 16.