Facebook scales back on chatbots was one of my previous posts. My conclusion was that it looks like were going back to the IRV menu’s when we called a bank.

But Facebook is not the only one who is guiding us in the chatbot evolution (I don’t think it is a revolution). Brice Berdah from Callr wrote an interesting post about The state of chatbot, from linguistics to ethics in his post The chatbot masquerade: creating a personality with NLP and grammar.


The central question is:
Are chatbots making up to their promises?

Natural language processing: the key to smart chatbots. Below a summary of one of my previous post ChatBots are coming….but how is NLU used? in which NLP is described more detailed.

  1. Tokenization is the process of breaking a stream of text up into words.
  2. Lexical Analysis involves the converting of words in a sentence in the grammatical nature of the words.
  3. Syntactic analysis involves isolating phrases and sentences into a hierarchical structure
  4. Semantic Analysis involves the meaning of each word
  5. Discourse Integration deals with how the immediately preceding sentence can affect the interpretation of the next sentence
  6. Pragmatic Analysis deals with context contributes to meaning

Example Lexical- and Syntactic Analysis:

So the technology is available but it is very difficult to get the chatbot understand the user for 100%.

This was about the technology, but why do we need chatbots anyway?

Chatbots need to solve a problem in a more convenient and straightforward way than for example websites or apps.

For the user of a mobile phone (and probably less on a desktop computer) a more convenient and straightforward is to stay in his favorite message app which is open all day on the mobile phone. The user doesn’t have to look for apps which run the task or open a mobile website.

Below some examples of a more convenient and straightforward way:

  • Make appointments, find time, place and more
  • Get the right information, on the right time, on the right place.
  • Make reservations
  • Make orders, the more casual the order, the better chat app fits for it
  • Find work at right moment, in the right place full filling your wishes of that stage of your live
  • Make payments, the more casual payments, the more it fits

Brice described in his article the seven sins of chatbots:

  1. Limited AI availability
  2. Are use cases really that strong?
  3. Lack of transparency: Who/Whom am I talking to?
  4. Lack of context awareness
  5. Lack of communications with existing business systems
  6. Lack of focus
  7. Lack of human escalation protocol

Summarized the 7 sins most important is that chatbot give the rights answers or predictions. It is the base and includes all the sins expect transparency. Transparency is very connected to behavior. Personally I believe as soon as bots are working well and it is transparent for the user if he talks to a bot, human or a combination, the user will adapt his behavior. To the bot the user will not create any empathy as it is not needed.

Short direct answers is very convenient for the user and very convenient for the chatbot as it is easier to understand.

Abd then the answer to the question:
The state of the chatbot, why do they still fail?
At first because the are not doing what they promise. Chatbots are not working well.
At second chatbots could be more focused on the problem to make tasks more convenient and straightforward
And at last, the user is not used to chatbots, it is not transparent and the behavior is not adapted.

We, at ContentForces, creating better chatbots. If you would like to know more contact me via the form.

Leave a Reply

Your email address will not be published.