Text Analytics and Natural Language Processing: Knowledge Management’s Next Frontier
NLP Is Not a Commodity
Another misconception is the notion that all NLP solutions are the same. While it’s true that most natural language recognition systems have comparable levels of accuracy, NLP is an entirely different technology. Recognizing words is a basic function compared to assembling meaning and intent. And there’s a lot of debate about the various approaches to NLP.
The two main approaches are statistical and symbolic. The former is exactly what it sounds like: You train the system on a huge corpus of data; it identifies patterns, generates a model, and then predicts the meaning of some piece of language based on probabilities. Symbolic is based on hard-coded
linguistic rules that are developed by people and taught to machines. At a high level, symbolic NLP seeks to teach the meaning of words to the machine, while statistical seeks to predict appropriate responses to inputs based on what worked for similar inputs in the past.
While there’s much debate about the merits of both, neither approach is sufficient on its own. A solution that blends the two approaches offers more flexibility and long-term utility than a solution that sticks with one approach. However, the simple best practice is to determine your NLP engine based on your needs and your data.
After all, it comes down to data. These systems train on data, they are fueled by data, and they generate enormous amounts of data. It is therefore essential that you are data-driven in your assessment and selection of any solution.
Ultimately, data is the resource that will drive business and economic development for the next century. It’s fitting then that knowledge management is at the center of this new frontier in business. KM professionals led digital transformation efforts as the world embraced the “information economy.” Now, armed with technologies such as text analytics and NLP, they are leading the charge into the intelligence economy.
Contact Us:
1-800-4VERINT
www.verint.com/engagement
www.verintblog.com
©2019 Verint Systems Inc. All rights reserved worldwide.