Technology Overview for Text Analytics
I just finished reading Gartner’s latest research covering text analytics. It seems to me like it’s their best study covering this sector so far. It is well written, comprehensive and a great tool for data analytics enterprises or business managers who want to gain in-depth knowledge on the strategic value of text analytics.
I found the research especially relevant because of its comprehensive, concrete list of use cases. The risk management section is a bit thin because it doesn’t cover specifically how text analytics furthers operating risk management but, in general, it is a very valuable tool. By skimming this list, it should become obvious to any enterprise executive that ignoring the business value that such a solution could bring would be bad management.
I would have liked to see a ROI framework for each case. As I mentioned in a previous post, I think that there are some misconceptions surrounding the lack of measurability of the ROI for text analytics solutions. A nice follow up piece could include the addition of these frameworks. We usually offer a detailed ROI for our customers, so it would be nice to see a respected analyst like Gartner offering the same to its customers.
Hopefully, Gartner will forgive me for copying a section of their research. The following is probably the highlight of the research as it adds perspective for all decision makers in an enterprise:
“The relative maturity of text analytics in the content world: Text analytics is the most mature technology in the content analytics space. In 2014, text analytics is passing the stage of Trough of Disillusionment in Gartner’s Hype Cycle reports (see “UnderstandingGartner’s HypeCycles”). Text analytics is expected to go into the stage of Slope of Enlightenment in the coming few years because it will become more prevalent and because of the opportunities to be realized. Drawing on the experience of early adopters, understanding grows about where and how text analytics can be used to good effect and, just as importantly, where it brings little or no value. The abundance of text data sources will make text analytics a priority choice for organizations.”
My final observation is related to the classification of the different approaches. While I was thrilled to see a clear definition of what statistics imply (complete lack of understanding and need for rich sets of training content), I noticed that Gartner didn’t include semantics as a separate approach to text analytics. The value of disambiguation, the availability of a domain-independent and concept-rich semantic network and the related long list of relationships among concepts are extremely important elements which make text analytics projects more effective and simpler to implement. I think it could have been very helpful to add a section which describes why not all the linguistic approaches are created equal. Again, prime material for the next release.