ChatbotCreator is a python package for creating chatbots(Including discord bots).
Use the package manager pip to install ChatbotCreator.
pip install chatbot-creator
First you need to create a json data file and save it as "data.json" in the current directory for creating the machine learning model.
example:
{"data": [
{"class": "greeting",
"patterns": ["Hi", "How are you", "Is anyone there?", "Hello", "Good day"],
"responses": ["Hello!", "Good to see you again!", "Hi there, how can I help?"]
},
{"class": "goodbye",
"patterns": ["cya", "See you later", "Goodbye", "I am Leaving", "bye"],
"responses": ["Sad to see you go :(", "Talk to you later", "Goodbye!"]
},
{"class": "name",
"patterns": ["what is your name", "what should I call you", "whats your name"],
"responses": ["You can call me Nacky.", "I'm Nacky!"]
}
]
}
here "class" is the label, "patterns" are the questions for the label and "responses" are the responses or answer for the label.
[NOTE]: You should create the "data.json" file like the given example where it should first have the "data" tag and then other tags explained in the example.
To create a simple chatbot, first you need to create the pickle data files
from ChatbotCreator import ChatbotCreator
main = ChatbotCreator("model.hdf5") # Model file name to save
main.createData("data.json") # The data file that you created. This will create the pickle data files.
main.createModel()'''This will create a neural network model and save it
with the name that you gave in the first step.
'''
After you have created the pickle data files and the model using the above code, you can remove that.
from ChatbotCreator import Run
run_model = Run("model.hdf5")''' Enter the model file name.'''
while True:
inp = input("Enter cmd: ")
if inp == "q":
break
pred, response, results, results_index = run_model.run(input_variable=inp) ''' specify the input variable through which you will parse in the input values.
'''
print(response)
# this will run the model
Here- "pred" is the predicted label, "response" is the response for the predicted label, results are the probabilities for every single label and "results_index" is the predicted label's index in "results"
If the chatbot is not confident about a particular question then it will save that question in a text file named "low-confidence-patterns.txt" in your current directory. You can use the queries in the text file to train the model again.
To create a discord bot:
from ChatbotCreator import CreateDiscordBot
discord_bot = CreateDiscordBot("model_file_name", "bot_token", use_wikipedia=True)
'''
model_file_name is the name of the model file, bot_token is the bot_token that you can get in discord and when use_wikipedia is set to True, it will send wikipedia results when the model is not confident about a particular question.
'''
discord_bot.run() # this will run the discord bot
To use different spaCy trained pipeline languages (to make your bot speak spanish or french) you may also specify the pipeline name:
discord_bot = CreateDiscordBot("model_file_name", "bot_token", "pt_core_news_lg")
then
run_model = Run("model.hdf5", lang_model="pt_core_news_lg")
You may find different trained models at: Trained Models & Pipelines
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
See the source code here /chatbot-creator/ChatbotCreator/
Currently, the CreateDiscordBot class has some issues which are:
1st issue) It can't be ran as a python notebook, it raises an error which needs help. It can only be ran as a python script.(class CreateDiscordBot)
2nd problem) When running the discord bot, it will function normally and send messages to the discord server for the queries but it will not return the predicted label to us.(class CreateDiscordBot)