Are you down with NLP?

A non-technical introduction to natural language processing

Natural Language Processing

We are not talking about Tony Robbins’ favourite self-improvement technique, neuro-linguistic programming. We’re talking about Natural Language Processing (“NLP”), which is a field focused on creating models that will allow computer systems to understand the meaning of human language. The most effective current NLP models draw on machine learning, deep learning, and various pre- and post-processing analytical methods to optimize the language understanding of a computer system. The most popular implementation of NLP is for conversational interfaces, or as they’re more commonly known, “chatbots”.

Chatbots

A chatbot is an interface that allows a human to interact with a computer system in a conversational manner. Many users find a conversational interface a much more natural way to use a computer system, as compared to using forms or search engines or some other method of communication.

Typically, a user will ask a question of a chatbot system. The chatbot will use an NLP engine (among other techniques) to understand the question, and then will match the question within a knowledgebase of questions and answers, and then provide an answer to the user from the knowledgebase.

The chatbot may also ask questions of the user at points during the conversation. The purpose of the chatbot may be everything from answering basic questions to collecting data to completing a workflow to facilitating the purchase of a product or service.

Since a conversational interface is a basic method of human-computer interaction like a browser or a smartphone app, the applications are virtually endless. If you have humans looking to interface with computer systems, a chatbot is always an option. It should be noted that chatbots should not be viewed as a replacement for other interfaces, but more as another option within an omnichannel strategy.

The most well known chatbots right now are voice assistants- solutions like Siri, Alexa, and Google Assistant. However, the overwhelming majority of chatbots in the field are chat/text chatbots, since the cost of building a voice chatbot at this time is prohibitive, and mostly left to the Fortune100 technology companies.

Chatbots are available on all major chat platforms, including Facebook Messenger, WhatsApp, and Skype. A chatbot can also be readily integrated into any of the major CRM platforms like Salesforce, MS-Dynamics CRM, or Zendesk.

Chatbot Applications by function

As mentioned there are virtually an unlimited number of chatbot applications. We’ve broken these sample applications out by function, since they could apply to virtually any vertical — all levels of government, healthcare, financial services, consumer packaged goods, etc.

Customer Service

One of the first applications of AI in the field was customer service chatbots. It should be noted that this area may include regular customer service, technical support, specialized help desk, or any of many other types of customer service. The customer service may be in support of clients, partners, or colleagues. So a customer service chatbot could be deployed as part of a website, an intranet, or an extranet.

Assistant

A low risk method for starting out with chatbots is to deploy the technology as a real-time assistant for your chat agents. Rather than the chatbot interacting directly with users, it provides recommended responses to the chat agent who has three options:

· Accept a suggestion as is.

· Choose a suggestion and edit it.

· Reject all the suggestions, and write their own.

Following an iterative training approach, the suggestions supplied by the assistant to the chat agent will improve over time. Benefits of a real-time assistant include an improved customer experience, lower agent turnover (due to increased job satisfaction), and lower costs (do more with the same number of agents).

Direct/After Hours

The most common deployment of chatbots is to interact directly with clients. This may be a first option, an alternative to waiting for an agent, or an option after hours when the contact centre is closed. Chatbots can reduce waiting times, improve the efficiency of a contact centre, and ensure appropriate messaging via standardized responses. As a chatbot is exposed to more questions, and continues to be trained, its ability to achieve first call resolution will improve. Of course, if a chatbot is unable to satisfy a user, they can always be directed to an agent (with the chat transcript sent to the agent for review).

Repetitive Tasks

Any chat agent will tell you that their days are often consumed by completing repetitive tasks like password resets, providing copies of receipts, and issuing RMA numbers. These types of tasks are well suited to chatbot technology. Any repetitive task with a standard workflow can be mapped out and incorporated into a conversational interface. Instead of having to wait for an agent, or fill out a form, a chatbot will ask the user for all the necessary information to complete the workflow, and will resolve the call without the need for any human intervention.

Promotion and e-Commerce

Qualification

There are many specialized chatbots that collect information from a prospective client before passing the client on to an agent. This can ensure that the client is routed to the right agent, and that the client’s time can be optimized.

Promotion

Combining a chatbot with a creative concept and copy can be a very effective method for promoting a particular service, product or campaign. The chatbot answers can even be chosen to provide it with a certain “personality”, be that extra friendly, officious, or sarcastic, for example. All kinds of content can be incorporated into the chatbot to create a responsive, interactive experience which can effectively communicate the underlying message of the campaign.

Since chatbot users tend to skew younger, a promotional chatbot can often be an effective way to communicate with a demographic that can be difficult to reach via traditional methods.

e-Commerce

Chatbots can be integrated with an e-Commerce engine in order to facilitate payment as part of the chatbot experience. After the chatbot has helped the user decide what product or service they would like to purchase, the chatbot can collect all the necessary information in order to complete the transaction.

Human Resources

The Human Resources department will typically receive three classes of questions from employees:

· Questions about policies and procedures (what is the hiring process? Can I work from home once a week? What will happen in my performance review?)

· Questions about benefits (is vision covered? Do we have travel insurance? Can I expense my course?)

· Personal Questions (How many vacation days do I have left? Have I maxed out my dental coverage? How many candidates have you found for my job opening?)

Many of these questions are well suited to a conversational interface. Your employees can get quick answers and your HR staff’s time is freed up.

Chatbots can also be used as a screening tool during the recruiting process and can also be deployed as an assistant to new hires during onboarding.

Integration

The value and effectiveness of a chatbot is often amplified by integration into your enterprise software. Integrating a chatbot into enterprise systems like an ERP or CRM, can greatly widen the range of questions that a chatbot can handle, and increase its autonomy. Given that virtually all enterprise solutions offer API’s for third party solutions at this time, this type of integration is typically not expensive or time consuming.

*BONUS* — Natural Language Generation

An area similar to NLP is NLG — natural language generation. This discipline involves creating text summaries of primarily numerical data. Some applications would include automatically creating a press release based on a company’s quarterly financial statements. Another application would be generating a game summary based on the box score of an athletic event. Like NLP, NLG has many promising applications.