Text Analytics – OK Sherlock, what’s next?
Text analytics sounds like fun, although most professionals in the industry probably would not agree with me. It’s like being Sherlock Holmes and finding the deep dark secrets held in our content repositories. The problem of course, is how are you going to find these deep dark secrets? You need a plan. Sounds rather dumb doesn’t it.
One of the steps that is typically overlooked is cleaning up your data. We find in most organizations that unstructured content and semi-structured content isn’t proactively managed. In simple words, it’s a mess. When you attempt to perform text analytics, specifically with content, it is garbage in and garbage out. Get rid of the dupes, the dark data, the content that should have been archived and wasn’t. Be very careful of personal or confidential information. If you are planning to run a marketing campaign with the results, you may step over the line.
Define what you want to do, either reduce costs or increase revenues is typical. From that generic definition, you can decide that improving processes is a way to reduce costs, or increasing repeat sales is a way to increase revenues. Next, define what information you need and then run a test on a small sample of defined content.
Use the human intellect to plan the project from beginning to end, and then just execute. Exactly. My dear Watson.