As we noted in the post “Chatbot: What is Chatbot? Why are Chatbots important?”, there are different approaches you can follow to develop a chatbot to improve your customer service applications.
In evaluating which chatbot is right for your application, you may consider several criteria, such as:
● Use case. Is the chatbot a customer service application or is it a back-office application? Or is it both? Should it only simplify and enhance the interaction with customers by offering information they need? Or should it allow customers to perform relevant actions such as direct access to wire transfers, deposits, etc. in the case of a banking application?
● Information. Is there already a structured knowledge base? Do you already have high quality content and examples that can be used to train the chatbot system?
● Technology. Chatbots are a disruptive force that have touched every industry in recent years. However, not all chatbots technologies are the same. The ability to understand the meaning of words and process language is a key differentiator.
Why words are so important… also in customer service applications such as Chatbots
Different technologies have become more efficient and able to effectively resolve an increasing number of complex problems. For example, there is still a lot of hype around machine learning but there is also an increasing awareness in the enterprise world around the limitations of machine learning (and deep learning) and an ever-increasing interest towards more intelligent approaches.
Machine learning (ML) in customer service applications. How does machine learning work? It’ basically pattern recognition. This means that machine learning algorithms work almost independently from a specific language. This may seem like an advantage, especially if you have a lot of content that can be used to train the machine learning models to perform pattern recognition. But the truth is that machine learning models struggle to recognize context and it goes without saying that if a system that aims to improve the interactions with customers through natural language understanding via chatbot does not recognize context, it won’t be able to understand a customer’s needs. As a result, ML-based chatbot systems may offer brittle and inaccurate results due the inability to identify and properly comprehend the different meanings of words.
Read more on machine learning! The truth about machine learning (and deep learning)
Applying natural language understanding (NLU)
The ability to understand a customer’s request and identify, extract and provide the right information coming from your knowledge base is the key to implementing really effective customer service applications.
As described in the following diagram, the “question understanding” module of a chatbot plays a key role: the more detailed and accurate the requests analysis is, the better the results.
Question Understanding processes based on Natural Language Processing (NLP) & Natural Language Understanding (NLU)–Artificial Intelligence software that leverages text analysis, linguistic and semantic understanding, etc.–aim to understand customers’ requests in the most accurate way.
Once user intent (in the form of a request) has been processed and analyzed, the Dialog Manager module retrieves the right information from the enterprise knowledge base (or knowledge bases) or by accessing the different Line of Business (loB) applications not only to retrieve the most appropriate information, but also to identify specific data the customer needs to perform a specific action.