As a technology decision maker, all the hyperbolic talk about having a cognitive solution that can converse with your customer might seem like years off. Understanding natural language processing (NLP) maturation coupled with the right uses, today’s NLP achievements can lead to significant benefits, cost reductions and consumer satisfaction.
One could contend that NLP maturation can be broken down into three distinct maturity levels: inquiry NLP, conversational NLP and reasoning NLP. Below are the descriptions of these maturity levels.
Inquiry NLP is the capability when you use plan English to find or ask for something. Typically, this type of capability deals with retrieving simple facts. “Show me all pink sweaters made out of cashmere for a petite size woman.” Or, “How much does a blue whale weigh?” This is the most basic level of NLP and reflects the ability to breakdown a sentence structure into entities: person, place or thing. This capability is accomplished with text analytic tools, but lacks the ability to perform co-referencing — connecting sentences together through context. This capability recognizes nouns and verbs against a well-known taxonomy or ontology. As an example, Schema.org is an open community internet ontology that is sponsored by Google, Microsoft, Yahoo and Yandex and is used by 2.5 billion web pages that defines concepts for searching across the internet.
Conversational NLP is the capability that starts to define context and user intent. This is sometimes referred to as natural language understanding (NLU). This capability allows the system to narrow a set of facts through “conversation” by the system prompting the user for clarification. “Siri, call my doctor.” Siri responds with: “Mitch, which doctor do you want me to call, Dr. Smith or Dr. Wesson?” Some of the major players in this space include Apple Siri, Google Now, Amazon Echo, Microsoft Cortana, IBM Watson and the latest to join the conversation list is Viv Lab, which was recently sold to Samsung. More advanced Conversational NLP systems like Viv Labs are now demonstrating capabilities to understand user intent. This is the capability that incorporates concepts as part of the narrowing of facts. Take the example of ordering pizza: “Viv, Order me a medium pizza.” “Mitch, what topping do you want on your pizza?”
The system is using the concept of necessity and sufficiency of relationships to understand what features are required to complete the task of ordering a pizza. The concept of necessity and sufficient exhibits the illusion that NLP systems are “thinking.” In reality, necessity and sufficient concepts allows a NLP system use conversation approach to narrow the facts. At the end of the day, no matter what the pundits write, conversational NLP is still a factoid machine and does not reason.
Reasoning NLP is the holy grail of artificial intelligence (AI). Reasoning NLP provide the ability to apply critical thinking and answer open ended questions like “What happens to a person who steps off the back of a pickup truck moving at 25 miles per hour?” The ability to reason not only requires understanding concepts like space, time, inertia and mass but, also requires understanding regional, cultural and religious beliefs. Without understanding regional, cultural and religious beliefs, an AI system could inadvertently offend or alienate the consumer base that your trying to reach. You don’t have to look any further than the pizza example above. Having an AI system offer meat toppings on a cheese pizza to those consumers who follow kosher guidelines would be inappropriate to say the least.
Having an AI system identify cultural dietary practices is one thing, what about the cultural dilemma that Susie Walker from the 1947 classic movie Miracle on 34th Street runs up against? Could an AI system be able to reason with some childlike innocents that Santa Claus exists? Would an AI system understand the concept of faith, having a strong or unshakable belief in something, without proof or evidence? How would an AI system be able to reason without proof or evidence?
Obviously, reasoning NLP has a long way to go in handling such enticing questions before it can become commercially feasible. However, this shouldn’t stop a company from engaging in the most preliminary use of NLP.
Let’s take a look at a hypothetical health insurance company that uses inquiry NLP to provide consumer empowerment. Inquiry NLP can be used as a concierge service to help their members in a number of areas, below are some examples of empowering consumers:
- Find a doctor: “Find me a doctor in my plan that treats shoulder pain.” The health insurer already knows a great deal about their member’s medical background, including which plan that are enrolled in, where they live and what hospital networks they prefer to use based on their previous claim usage. The system would present the member with a list of top doctors that participate in their plan and are geography located close to where the member resides.
- Select a plan: “I’m married with two kids and I like the freedom to select my own doctor, which plan is best for me?” The system should be able to look at your age, past claims and medicine that you and your family are currently using to develop a predictive usage for the next enrollment year. Based upon those predictive usage patterns, the system could make a recommendation for which plan would best fit your profile.
- Summary of benefits: “How much will I need to pay out of pocket to see a shoulder specialist?” The health insure will already know the amount of out of pocket money you have spent and how much is left in your deductible before the insurance will start to cover your expenses. The system should be able to give you a summary of costs for any particular procedure or service that is performed under your plan.
A customer shouldn’t have to navigate through a maze of web pages, or get hold of a customer representative to find actionable insights. The goal using NLP is not to replace your customer service representative, but to empower your customers with abilities to find real value immediately.
An incremental NLP strategy could be used to augment the existing customer service staff that are employed to handle such questions that they receive from their members. The first stage could start off with an inquiry NLP system that would answer simple questions that a member types in plain English. If the system doesn’t know the answer, the customer service staff could type in the answer, training the system for the next time the question is posed. The second stage is to integrate a conversational NLP that has a good grasp of enterprise concepts that allows the system, through prompting the user, to narrow the facts as in the case of performing plan selection and “which plan is best for me?”
NLP is the tool that requires focusing on the development of knowledge. Knowledge is the instrument that enables your customer to navigate through ambiguous web of complex terms and concepts. One capability of NLP that has not been written or discussed much, is the ability of an NLP system to be able to write its own computer code. Most business applications are about retrieving, reporting on data or performing some type of workflow. It is not inconceivable that at some future date, an NLP system will understand well written business requirements. Such a system would be able to generate its own queries or workflows against today’s platforms to perform information retrieval, reporting analysis or develop workflows to meet the business needs.
To compete in the 21st century, companies will need to apply NLP to meet the every growing consumer expectations. This will require developing a strategy for harvesting knowledge to realize true benefits. Understanding the need for a strategy and the development of knowledge acquisition systems that inter-operate with natural language will be the competitive advantage that every company will need to achieve in order to keep and grow their customer base.