Chatbots are becoming increasingly common on websites and in messaging apps. They are used in customer service, marketing and sales. Before integrating a chatbot, companies should ask themselves how the chatbot will interact with users.
There are essentially two types of chatbot:
1. AI chatbots that allow free-text input and are therefore usually based on natural language processing
2. Rule-based chatbots, which offer fixed click structures so that the user cannot enter free text but instead ‘clicks through the conversation’ using buttons.
These rule-based chatbots are not based on AI, whereas chatbots that can understand and process free text using NLP do so with the help of AI.
What exactly is NLP? NLP stands for Natural Language Processing and enables machines to understand written or spoken human language. What may seem simple at first glance is actually a highly complex technical undertaking. For even though it may seem obvious to us – understanding words and semantic structures is not easy for machines, as natural language and grammar are far more complicated than one might initially assume. In addition to the words themselves, the meaning and context of the sentence must be understood, which can be very complex given the ambiguity of natural language. Unlike humans, computers cannot draw on experience and past events – or so it would seem. Thanks to AI and machine learning, however, computers are becoming increasingly intelligent. The aim of NLP is to enable communication between humans and computers. Key areas of application include, for example, communication between humans and chatbots, or between humans and digital assistants such as Alexa or Google Assistant.
Natural Language Understanding (NLP) vs. Guided Chat Dialogues
The strengths and weaknesses of both types of chatbots mentioned above are listed below based on many years of experience and practice in the area of chatbots.
NLP Chatbot: Finding the Right Answer Using Free-Text Input
The free-text input field allows users to type their queries directly into the chatbot. Using state-of-the-art NLP technology, the chatbot analyses the input and assigns it to specific ‘intents’ (query types). If the chatbot has a suitable response for a particular intent, this is sent back to the user.
The big advantage for users is that a concern can be reported quickly and easily.
However, it is difficult for the company to answer every question correctly from the start. With every user input, there are three possible results:
1. The chatbot correctly understands the user and gives a correct answer.
2. The chatbot misunderstands the user and gives an incorrect answer
3. The chatbot doesn't understand the user at all and cannot send back an answer.
The last two scenarios quickly lead to frustration on the part of the user. Only in the first scenario is the user experience (UX) positive. However, the last two scenarios do offer an advantage for the company using the chatbot. They learn what their users or customers like to ask or want to know from the chatbot and can, if necessary, respond correctly to such queries in future or adapt existing intents and expand the answers. Incidentally, with moinAI, this happens automatically; in other words, the AI independently clusters the questions it has not understood and suggests to the relevant companies that they create topics for these issues. This AI feature is called Dreaming.
The following table shows the advantages and disadvantages of the NLP approach based on a SWOT analysis:
Click-Based Chatbot: Guided Conversations and Buttons
The second approach to providing users with the right answer to their query involves guided dialogues. Buttons or so-called quick replies attached to messages guide the user to the desired outcome. The buttons may suggest topics (e.g. reset password) or, in the case of quick replies, offer a choice between “Yes” or “No”. Depending on what the user selects, they receive a response from the chatbot.
Although the user’s freedom is limited, inputting via buttons allows the chatbot to interpret the user’s queries more accurately. This results in a more meaningful response. Good conversational design can quickly guide the user to the right destination and draw their attention to further content. However, the use of buttons limits the user’s freedom, and fewer or no text messages are sent to the chatbot. The downside of this is that new intents cannot be captured. These are the strengths, weaknesses, opportunities and risks of a click-based chatbot:
Checklist for the Advantages and Disadvantages of Both Types of Chatbots
The following table shows the advantages and disadvantages of the two types of chatbots:
Chatbot Types: Which One Should you Choose?
It is impossible to say definitively whether a rule-based click bot or an AI-based chatbot – which allows and understands free-text input – is better. This is because it always depends on the objective the bot is intended to achieve and what the company expects from the chatbot. A click bot is ideal for simple and seasonal campaigns or marketing promotions. The clear advantages of click bots are that they do not need to be trained and are significantly simpler in design than AI chatbots.
However, if companies are looking to implement a long-term solution, for example in customer service or marketing, it might make more sense to invest a little more time and resources in implementing an AI chatbot. Because there is one thing we can definitely confirm based on our many years of experience: the time and effort invested in the initial setup of an AI chatbot pays off within a few weeks or months. Once the AI chatbot has been trained and the response texts formulated, the chatbot continues to develop independently thanks to AI and machine learning.
Nevertheless, it should be noted that an AI chatbot that allows free-text input naturally always carries the risk that some user queries may not be understood. But this weakness is also a strength: every question it fails to understand is valuable to the chatbot, as it learns from them. You can also safeguard against this by using so-called quick replies, i.e. buttons. This means that even if the chatbot has not understood a query, it can still offer the user further options, such as forwarding them directly to a human member of staff.
.png)
