The rise of chatbot technology is quickly proving to be more than a mere fad, with a number of major businesses—such as Amtrak and Marriott—enjoying significant financial benefitsfrom their use in customer service settings. The ability of chatbots to effectively engage with customers is due to NLP (Natural Language Processing), a technology that enables computers to understand elements of human language. If you want your bot to understand questions posed by your customers, then NLP is a necessity.
First, though, what is NLP? Natural Language Processing allows computers to derive meaning from user text inputs. Unlike a search engine, which simply acts on keywords, NLP uses knowledge of sentence structure, idioms, and machine-learned pattern recognition to match user input to an “intent.” All intents are classified, which simply means that a desired chatbot action is associated with each intent. For example, asking a chatbot for next week’s weather forecast in Dallas results in the NLP tech attempting to understand the textual question, correlating that with what the user wants (user intent), and then forming a grammatically-correct response.
NLP technology is currently in its infancy, with an average precision rate of around 60% – 70%. More often than not, an NLP-powered chatbot will understand what you’re trying to ask if conditions are satisfied (i.e., it can understand the question and it can match an intent with an action). There is still a 30% – 40% chance that the chatbot—depending on the situation—will either provide an irrelevant answer or simply not understand the question.
Comparison of Chatbot Use Cases
NLP is an important ingredient in the world of human & computer interaction, though its efficacy depends largely on how and when it is applied. At a high level, there are three common use case scenarios for effective chatbot usage:
Chatting with Options
This scenario is characterized as a chat that has a specific purpose with a limited number of potential options. Examples include ordering take out food, shopping for new clothes, and initiating a customer service request.
All of these interactions are relatively focused, enabling the chatbot to retrieve data that is relevant to the customer’s inquiry. After a question is asked, the customer will be presented with options that, when selected, essentially drill-down into more detailed selections. Customer data can also be asked and recorded during the chat, with NLP playing a role in humanizing the chatbot, especially during the initial opening & closing of the chat sequence.
Common Scenarios: Service requests, case management, technical support, shopping.
Pure NLP-powered Chatting
This type of chatbot engages in direct question-and-answer interactions with a user while relying heavily on NLP technology to understand the intent of the questioner and providing relevant answers. Ideally, machine learning technology is also used to improve relevancy over time.
Realistically, this type of chatbot is feasible when there are some limitations placed on the scope of the topic. For example, a chatbot-powered virtual assistant could provide relevant information when integrated with customer-specific data sources, such as a CRM, calendar, e-mail, etc.
Topic limitations are needed due to the current maturity of NLP technology — human conversation is one of the most complex cognitive actions in the world. Asking even the most advanced bot to understand the intricacies of language, particularly irony and sarcasm, is a bit too much to ask of any microchip.
Common Scenarios: Chatbot assistant, shopping, banking, reservations.
A chatbot variation that is currently gaining traction is the Conversational Interface (CI). In this scenario, the roles are reversed and the chatbot is the one asking questions. By putting the proverbial ball in the court of the responder, the chatbot is able to control the flow of the conversation by leveraging user-provided replies to humanize the interaction. More often than not, options are presented to the user based on previous answers.
The result is a highly personalized experience that more closely mimics real-world conversation. With CIs, Natural Language Processing and Machine Learning may not be used — it is largely comprised of current Web technology, which results in lower costs.
The downside of Conversational Interfaces is that conversations are only feasible in certain situations. A customer who wants to lodge a complaint is not interested in circular conversations with a myriad of paths — they want action in the form of data provided and received.
Common Scenarios: User experiences, tutorials, eLearning, conversation chatbots.
Natural Language Processing is still a work-in-progress, but the fact that the current technology is still viable enough to positively impact the corporate bottom line is a testament to the promise of NLP-fueled chatbots in the future.
The future of NLP and chatbots may shine brightest in industries where customers require immediate engagement with companies. Customer service solutions, IT help desk, and service request platforms can all benefit from the instant attention—and access to customer data—that chatbots can provide.
In these industries, companies are moving towards hybrid solutions that combine the data-gathering skills of chatbots with existing support & customer service personnel. It’s not uncommon for organizations to implement chatbots as their preferred frontline Level 1 support, with escalations being sent to human caseworkers.
As the technology continues to improve, specialized chatbots may be able to fulfill more advanced roles as support technicians work to refine and improve customer service workflows.
Crow Canyon Software delivers the power of a connected, digital workplace to your organization, helping you be more successful and competitive. We provide the tools that allow you to gain the maximum benefit from SharePoint and Office 365, Microsoft’s premier collaboration platforms.
Want to learn about how our business productivity applications for SharePoint and Office 365 can power up your organization? Give us a call at 1-925-478-3110 or contact us by e-mail at email@example.com. We look forward to hearing from you!