Natural Language Understanding and Conversational Dialogue - A Different Kind of Self-Service Speech Recognition

By Stefania Viscusi

TMCnet Assistant Editor

 

Speech recognition technologies, with the ability to automate and add voice to various applications, are becoming increasingly intelligent--making natural conversation between human and machine possible.

 

This conversational element gives users more power and freedom—further improving interactions and offering increased accessibility.

 

By allowing users to ask questions rather than having to find a category to fall into, it is possible to save money and complete more transactions in less time.

 

For more insight into natural language understanding in speech technologies, I took some time to ask Luis Valles, Chief Scientist at GyrusLogic, some questions on the topic.

 

Natural Language Understanding (NLU) or Conversational Dialogue is the capability for a user to say and/or ask anything, and the system understanding what the user meant, together with the system finding an appropriate response—as with any other conversation between humans. 

How is this deployed with Speech Recognition?

NLU can be deployed in different ways.  The traditional method of deployment is by a system which is based upon Statistical Language Modeling (SLM) and Statistical Semantics Modeling (SSM).  To initially implement this type of solution, there must be approximately 30,000 transcriptions of potential phrases to be recognized.  These phrases will populate the statistical model, and based upon those statistics, the system will start determining the likelihood of a word combination from its frequency in the populated model. 

The GyrusLogic approach is very different because the basis of the system parallels the basis of language—grammar, context and semantics.  Our engine receives a phrase from any Speech Recognition engine—a VoiceXML (News - Alert) browser, IVR or web chat session—which needs to be answered; and it will understand the meaning of what the user wants and act accordingly.  The system will recognize if the user wants to execute a transaction or simply wants a question answered; either way it will give the user an answer or will ask for potential missing information related to the execution of a transaction. 

Do you believe that moving to a more Natural Language Understanding with speech will change the ways in which people are using speech today?

Yes, the NLU or Conversational Dialog capability will definitely change the dynamics of automated self-service.  The big difference will be the user’s ability to ask anything they want, without being forced into lengthy menu structures.  Users today are getting more and more exposed to this type of capability, and in a few years, people will not know any different, this will drive PC’s with voice, as well.  This should occur by noting the evolution of self-service solutions: in the 1980’s people started using touchtone IVR’s, in the 1990’s the web text channel was incorporated, in the late 1990’s we started using directed dialogue Speech Recognition and we will see NLU self-service as a logical next step.   

How does natural language understanding answer to the troubles of long menu trees? 

With NLU applications, the caller will be in charge of the dialogue and asks questions, in their own words, for the computer to generate responses; in contrast to most of today’s applications, where the system directs the dialogue and the caller continues to get re-routed until arriving at the appropriate destination to get an answer.  Thus, the NLU basically eliminates menu trees, creating dialogue on the fly since one never knows what the caller will say and/or ask, giving the caller responses much quicker.

Does this offer cost savings, if so how?

The cost savings for NLU applications can be found in several ways, specifically, there will be a significant savings in development and deployment costs since we introduce a paradigm shift with development from traditional procedural code developments to a declarative methodology based upon business knowledge and information—potentially saving 30% in development time.  Additional significant operational savings can be achieved due to the fact that people will spend a lot less time on the phone.  Together with less time on the phone, the potential for less concurrent licenses to be purchased can be a substantial cost savings.

What are some of the benefits a business can recognize from utilizing this kind of speech?  

The GyrusLogic technology will assist in a logical improvement of fixing speech recognition false positives since it’s based upon language, context and semantics.  The other benefits, as we discussed before relate to total cost of ownership, which will dramatically decrease.

Can it improve customer satisfaction?

Benefits are found in improved customer satisfaction since the user will get an answer right away without many menus to go through.  In other words, first contact resolution is a fact, but achieved with the customer in control with less time on the phone.  It is a very different question to ask the customer “How can I help you today?” and the system is able to handle it without lengthy menus.

When it comes to self-service, how can natural language offer improvement? 

The self-service can benefit in many different channels with this type of solution, since our system works on a text basis, allowing utilization of the same application framework for different channels.  This allows the customer to receive a consistent answer from their self-service question—no matter which channel is chosen.

Do you think there is still a long way to go to have pure person-computer communications using speech?

The infrastructure and technology is ready today, and is a logical requirement of users moving forward with their demand of improving functionality of computers.  We believe in the next 18 months, natural language understanding will be a standard requirement, as the demand from users already exists.

What are some examples of the applications where natural language is currently applied to or future places where it can be used to better a system? 

Most of today’s self-service applications are serious candidates to incorporate natural language capabilities, creating an improved way to customer experience.  Typically, applications are generated, or will be generated shortly where users simply ask product and/or service questions.  Other applications will be created for when users need to order products or services, and have questions at the same time.

Can you tell me a little about the solution you have developed to provide Natural Language Understanding capabilities? 

 GyrusLogic’s Platica product is patented artificial intelligence (AI) technology built with computational linguistic models for customers, employees or any other stakeholder to enter into a fully-automated conversational dialog.  A GyrusLogic Platica Self-Service application understands what the user means, regardless of how the question or request may be phrased.  The system’s answers are consistent and complete, and are returned in text and/or clear audio.   The following are examples of the high-level intelligence inherited in every GyrusLogic application—understanding complex natural language sentences, not requiring any specific additional application development:

 

The user can make corrections at any time in natural language, with the implicit correction capability

The ability for users to interrupt a transactional dialog with a non-related question or request.  The system will answer the question and will keep track of the dialog context and return to the original dialog, if needed.

The user can be spontaneous with any of their requests or questions at any point in time, allowing users to be specific or vague, or state open-ended questions.

Also, the developer can connect and disconnect to our engine at any time, making it easier to get natural language introduced in an organization.


For more information, click HERE