Sin categoría

What is a cognitive chatbot and how does it operate?

9 de September de 2021

In today's market there is a wide offer of Chatbot solutions as a virtual assistant that can do almost anything. Nothing could be further from the truth! There are free platforms, paid platforms, agencies and consultancies that offer this type of services and all of them have different or similar monetization models and scope of their services. In this article I intend to let you know a little more about a concept that is key within the objectives that you have as a company or implementer of this type of tools. This is because as there is both long and wide on the subject; in many of the occasions you can be short of information to make the right decision when it comes to knowing which Chatbot agency to work with or what kind of software to use to deploy your Bot. To begin with you should know that there are Chatbots that do not use artificial intelligence (AI); these types of Bots are used to solve some FAQs of your users such as: User says: Where are you located? Bot answers: Calle Joaquín Turina 2, Oficina 6 - Pozuelo de Alarcón Also this type of Bots are very used to obtain data or information that the user gives you within the conversation, for example: Bot says: What is your name? User says: Pedro The data that captures the Chatbot would be *Pedro* which is equivalent to the name of the user within the conversation. So far, artificial intelligence is not used in this Chatbot. On the other hand, there are SmartBots or Cognitive Chatbots, which use artificial intelligence to respond more accurately to user requests and/or queries. But ... How do they work? Cognitive Chatbots use NLU (Natural language understanding), i.e. compression of the natural language used by human beings. They use within their structure the interpretation of the contexts of a human conversation and apply Machine Learning (they learn from each interaction with users) to become more effective when giving an answer. Let's go deeper! A cognitive bot begins to interpret the sentences or audio sent to it by its users, by means of intentions. This intention is nothing more than what the user wants to do or receive in return with the message sent, for example: - User says: I want to know my account status In this sentence the NLU interprets the intent of the sentence and brings the user into context of the situation. For the example in question the Chatbot could proceed to say something like: - Bot says: I need you to please confirm the password with which you make web queries in order to provide you with your account status information. The request made by the user can also have input data that enrich the conversation to resolve the intention and make it more robust as in this case: User says: On June 01 a debit appeared in my account for USD35 that I don't understand what it corresponds to. The data (date and amount) provided in the context of this sentence are called entities. How to have a Cognitive bot? The first thing to do is to define the intentions that the Chatbot will be able to answer and the entities associated with that intention. In the case of the example above, the intention would be to review movements and the entities would be date and amount. The next step is to train our Chatbot with phrases, words or similar expressions that are in line with the intentions and entities. For example: I have a debit of XXXXX from the past XXXXX and I have no knowledge of it. This intention is the same as in the previous example. Entity 35 Dollars, 35 US Dollars equals USD 35. As a third step, we need to create the Chatbot logic. What should our Chatbot respond when a user triggers some of the intents. How are we going to respond to the user. Last but not least, we must take into account integrations with different (CRM, Web services, ERP, etc... ) systems where the bot will be able to query relevant information for the user and be fully transactional. Deploying the Chatbot With the previous steps completed, the next step is the deployment of our Chatbot. There the learning of the tool is semi-automatic; this is because when the Chatbots receive intentions from the user and are not sure of what they interpret, they notify the technical team of the same to add or give management to these intentions and thus strengthen the knowledge base of the Bot. Finally, for a cognitive Chatbot to improve its response rates and become better and better, it must be in constant training. This continuous learning involves interdisciplinary groups from marketing, communication and linguistics that contribute to the improvement of the conversations and the user experience with the tool. I hope this kind of information is helpful and valuable to you. Contact me if you need help or want to share some experiences with this kind of tools. Best regards, Javier Velásquez.

Loading...