skip to Main Content

I Want to use a website chat in our product sales page.
requirements are
1. this should have an artificial intelligent chat bot who is capable of answering the basic questions.
2. and admin panel from where my employees can monitor the chats as well as respond to customers?

is it possible?
what will be the best platform for it?

2

Answers


  1. First question “is it possible?” – Yes, it is possible.

    Second question is far more complex and you need to do some investigation before choosing best tools – depending on how you want to tackle it and what are your use cases. You can take a look at those articles for nice start and this comparison which I have found helpful while building my chatbot. Regarding this artificial intelligence – it is not a tool which will create all the conversations for you – you still need to do lots of typing but those tools will help you to match what user just typed into one of few intents which you can use to create dialogs. While choosing NLP solution please take a look at supported languages, sdks (if you need to enrich data with your backend responses) and if you care a lot about privacy of your users at legal notes. Some of them are for free and for some of them you need to pay. Some of them gives you different set of tools and some of them really basics and then you need to build something on top of it.

    Login or Signup to reply.
  2. Yes, it is definitely possible.

    There are two parts to chatbot in your case

    • Intent Identification
    • Retrieve/Generate Answers

    Intent Identification is part of NLP, in short it is text classification where you group the queries/questions within classes ( like PRODUCT,HR,SALES etc ) which visitors to your page are likely to ask. Then you basically train a model to learn the questions against each of those classes.

    Once you know the class of a question you have two options to retrieve the answer,

    • one is to do a simple lookup against the question and fetch the per-determined answer, here your question will never be the same so lookup will almost always fail unless you preprocess your question to some fixed format.
    • Second, you can have more intelligent approach by implementing a ML model.
      Model can be either be using TD-IDF or it can be complex seq2seq or retrieval model.

    These model can generate corresponding response against a question or can match best possible response against a question. You have to pick up response with best score and return it to the chat window.

    For Intent identification rasa-nlu is easy to use.

    For response generation you can check rasa website for rasa-core or you can pickup appropriate model from tensorflow/keras and get it working.

    you can check my repo for some sample codes – https://github.com/sarweshsuman/chatbot-codes-collection/tree/master/rasa-core

    Or you can go for something like – http://heyday.ai/

    Login or Signup to reply.
Please signup or login to give your own answer.
Back To Top
Search