ServiceNow adopted intelligent automation solutions called Now Intelligence to automate the service delivery process and scale service delivery efficiencies, while generating personalized experiences to users. ServiceNow offers numerous advanced AI-driven automation solutions, like performance analytics, predictive intelligence, virtual agent, and agent intelligence. To accurately predict a user’s intent and act appropriately, ServiceNow offers a Natural Language Understanding (NLU) module.
What is Natural Language Understanding?
Natural Language Processing (NLP) enables the system to read, analyze, interpret, and understand intent. Similarly, the NLU component analyzes strings of text to decipher meaning and intent. NLU is part of an NLP system that uses software to understand user inputs in text or speech formats. AI-enabled NLU models are trained to predict and understand statements or conversations and relate them to an intended task. The application using the NLU model can execute the user request. NLU is what lets chatbots understand and deliver natural and engaging conversational user experiences.
Challenges of Understanding Language
- There are multiple ways to describe the same thing
- Expressions can be indistinct
- Contextual information is essential
- Words can have new meanings over time
- Buzzwords, short-form words, and idioms can be complicated
The ServiceNow NLU Module
The ServiceNow Platform has robust Natural Language Understanding (NLU) so it can understand and answer a user’s intent. NLU was introduced in the ServiceNow New York release, but with ServiceNow Orlando’s release NLU model is equipped with full-scale integration of machine learning and agent intelligence capabilities and the NLU application is extended to other modules within the Now platform.
ServiceNow Orlando has many new NLU features:
- Users can use existing NLU models without cloning the NLU model
- Create custom derived entities based on needs instead of predefined entities
- Users can make entities available across NLU intents except in predefined entities
- The interactive dashboard measures the performance of NLU models
How the ServiceNow NLU Model Works
ServiceNow NLU consists of a model builder and an inference facility to help the system understand and react to user intent. Using a model builder, businesses can build the NLU model to their specific business requirements.
Figure: Process for Creating NLU Model in ServiceNow
Registering natural language instances in the NLU system lets the NLU recognize keywords and contexts of the user request. An NLU model is a mathematical representation of words, intents, and entities. The intent represents a user action.
Examples of Intents:
- Add emergency contact
- Check incident status
- Set up meeting
Entities characterize the context. They define a class of objects, with values representing possible objects in that class. Here are some examples of class and their values.
- System Entities: date, time, duration, location
- Glide Records: machine names, asset classes
- Company/Domain Specific: meeting rooms, company policies, industry terms
ServiceNow provides an NLU inference service to assist the NLU to recognize the language and trigger the appropriate activity. For any given user utterance, the NLU inference service coaches and calculates intents and entities based on data in the NLU model. Based on the action associated with the user text, the appropriate action is triggered into machine decodable formats.
Role of NLU in Virtual Agent
Chatbots are the most popular applications in automating various communication service deliveries. A single chatbot can handle many requests in natural language. NLU is what enables Chatbots to understand and respond to the user. To make chatbot or virtual agent conversations more human-like, ServiceNow leverages technology like NLU, which is integrated into ServiceNow’s Virtual Agent. This way virtual agents can enhance efforts to modernize various business processes. Virtual agents can easily:
- Raise tickets
- Provide a list of approvals
- Approve/reject change request/service request
- Create incident and service request
- Gather Knowledgebase articles and tables info
- Show dashboards
About The Author