Deploying text analytics and natural language processing for strategic advantage

Speech recognition 

One of the potentially more lucrative use cases for text analytics involves speech recognition, particularly when it’s based on exchanges between organizations and their patrons. “Enterprises have realized for a long time the value of understanding the voice of the customer,” Foderaro said. “You, as an enterprise, are doing yourself a disservice if you don’t start analyzing spoken conversations between customers and companies.” Text analytics is applicable to speech recognition once company interactions with customers are transcribed. Then, those spoken exchanges are subject to the same forms of NLP used for any other sort of text analytics deployments. In some instances, this might involve identifying meaningful words (perhaps about competitors, products, services, or business objectives) and analyzing how they were used in relation to defined goals. 

Examples of outcomes organizations have achieved with this sort of text analytics include being able to identify the most important questions asked by the customers and their main objections, Foderaro said. “Which products did they talk about and what competitor product? What are the main objections to a sale?” This method is aided by linguistic reductions, which are a rules-based form of symbolic reasoning that effectively remove all unnecessary words in sentences to reveal their bare minimum meaning. Thus, organizations can understand what their customers are saying in relation to defined outcomes (which Foderaro called “insights”) from text analytics. This can reveal vital information to organizations about what customers really think about them—and what they can do to monetize that knowledge. 

Human involvement 

Regardless of which approach to NLP (connectionist, symbolic, or a mixture of the two) is employed for text analytics, human involvement is an integral aspect to success. For traditional rules-based taxonomy undertakings, humans are necessary to engineer the knowledge at the core of the approach. For machine learning efforts, humans are often required to create the labels for training data and oversee the training processes in other ways. 

“Machine learning can nicely complement the human-in-the-loop defined principle, the design pattern for knowledge modeling in general,” Blumauer said. “Human-in-the-loop also means AI becomes more explainable and, also, to a certain degree, more responsible.” Even when using machine learning for automatic classification, extraction, and analysis, it’s critical to involve the human oversight of the process to ensure results are as desired. 

Making life easier 

By pairing certain elements of traditional rules-based NLP with contemporary developments in advanced machine learning, text analytics is getting easier for enterprise users. It also takes much less time to implement while readily diversifying an assortment of use cases for which it delivers tangible business value. NLP is one of the foremost areas of innovation in AI; text analytics and its users are assuredly the beneficiaries of this fact.