Chatterbot: How does the chatbot learn?

[juanpialbano]

Hi,

When the chatbot starts to work and answers the questions of different users, how does the chatbot learn?

How does the chatbot get feedback?

How does the chatbot know if the answer was correct or not?

Also

What are the best practices for building a chatbot? How should it be configured?

Should I use Adapters, Filters, etc?

[vkosuri]

@juanpialbano A detailed information available here http://chatterbot.readthedocs.io/en/stable/#how-chatterbot-works

However i would like add some information here on learning process.

1. Chatterbot is Retrieval-Based Bot , all retrieval-based bot are uses some pre-defined responses. If you want you could look chatterbot multilingual dialog corpus to get more families with chatterbot corpus.

2. After that you have to train your bot.

3. The training process stores all statements and responses into a specified database.

4. And also forms a statement-response-relationship

5. When ever user inputs statements, the response selection process uses some logic adapters to select proper response by compare each statement with inputted statement.

Please let me know if you are still looking any more information.

[juanpialbano]

@vkosuri thanks but I have doubt, How does the Chabot learn ?

here it doesn't explain so much : http://chatterbot.readthedocs.io/en/stable/#how-chatterbot-works

When the Chabot makes a response, it stores the request-response in the database for future queries?
How does he know if a request-response was right o wrong so he can use it in the future ?

Thanks in advance

[vkosuri]

@juanpialbano Thanks for your inputs we will improve that phase

When the Chabot makes a response, it stores the request-response in the database for future queries?

Yes,

How does he know if a request-response was right o wrong so he can use it in the future ?

Using Response selection methods

Let me know if you need any further help

[telkomops]

Hi Query on the response relationship. does the statement have user affinity. for e.g. User A Says Statement 1, User A Says Statement 2 , User B Says statement 3.
Stored as Statement 1 , Statement_2 ::: In_response_To Statement_1 , Statement_3 :: in_Response_To_Statement_2.
In here the user affinity is lost , correct? how is it typically handled by other users, do you format , 1 and 2 as 1 statement , so that indexing happens as Statement1+Statement 2 , Statement3 :: in_response_to_Statement1_Statement2 ?

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.