New tools help to unite unstructured data with what already is known to glean more complete views of customers.
Posted Sep 28, 2005
Speakers at Tuesday's Business Intelligence conference in New York City discussed how to make sense of unstructured data, including the use of the right tools with which enterprises can avoid potential customer disasters. An organization's ability to leverage the true value of all its data, according to representatives from SAS and Attensity during their presentation, "Discovery Through Data and Text Mining," relies on text mining, a spin on analytical processes.
"Text mining means different things to different people," said Manya Mayes, text miner product manager for SAS. To many it means search, "but search is a goal-oriented exercise--you have to know what you're looking for. You could still miss something that could be the demise of your company. Text mining helps you find new discoveries and prevent potential future problems before they become widespread."
Companies have spent a great deal of time and effort trying to understand structured data, so the key is not to ignore that information, but combine it with unstructured information to discover new things. Like any information-gathering process, however, Mayes warns that quantity does not equal quality. Just like data mining, a lot of effort can be spent cleaning. Some of the problems that surface with text mining include acronyms, misspellings, punctuation, and shorthand. She cited an example of the word dealer, which in one case was spelled 35 different ways, including combining it with other words. The solution is to automate the cleanup process.
David Bean, cofounder and CTO of Attensity, compared text mining to an electrical power grid. Using the right current allows people to send power across the country, instead of using a direct current from a battery, which won't get far. "Rows and columns make up the infrastructure that exists today. Data has to be in that form to do anything with it," he said. So companies must learn to turn text into rows and columns. "If I want to understand who did what to whom, when, where, and how often, I need to understand the linguistic relationship. If people didn't like eighth-grade English, this is painful to them."
Bean also explained how companies can diagram and dissect sentences into a fact relationship network with tools meant to deal with poorly formed information, such as when call center agents drop vowels or leave out words. "We can take a parsed sentence and extrapolate facts related to it immediately, pour [that] into a data warehouse, and use an existing analytic tool set to pour through data that came from text," he said. "This lets you see occurrences in your data you didn't know you were interested in until you saw it. It also lets you ask questions of your data you didn't think of before without having to reprocess that data."
"Given the opportunity, your customers can and will give you a lot of feedback on your products and services," Mayes said. "Text mining helps you combine structured data and unstructured data to get more information about your customers."
12 Tips for Generating Rich Data
Attensity's new Market Voice solution strives to harvest customer insight from both direct and indirect channels.
The three-company roll-up aims to enhance customer service by using voice-of-the-customer feedback.
Attensity's latest release incorporates insight from social networks to enhance a company's own knowledge base for better customer service.
With a goal to "open up the enterprise," the semantic analytics company purchases the global listening solution provider for an undisclosed sum.
Sponsored By: Informatica
Sponsored By: Verint