Armchair-expert is a chatbot inspired by old Markov chain IRC bots like PyBorg. It regurgitates what it learns from you in unintentionally hilarious ways.

• has been updated to work with discord changes
• added minor changes to the discord interactions
• Filters punctuations away for better or worse, it won't respond to random punctuations
• is a private bot, which you will need to run your own instance
• you will need to add your own data-set
• some specific packages you need to install, for the bot to work properly

• Uses NLP to select the most optimal subjects for which to generate a response
• Uses a Recurrent Neural Network (RNN) to structure and capitalize the output, mimicking sentence structure and capitalization of learned text
• Learns new words in real-time with an n-gram markov chain, which is positionally aware of the distances between different words, creating a more coherent sentence

• 3+ GB of RAM
• python 3.6+ minimum
• keras (Tensorflow backend)
• spaCy 2.0.0+
• spacymoji
• numpy
• tweepy
• discord.py
• sqlalchemy

I really recommend you install this in a virtual env, as you will install specific versions of some packages which might conflict on what you have. If not, you can just install it normally.

• This was ran under a Windows enviroment, so python could be python3 in linux. First: Run the Requirements.txt file in the root folder:

pip install -r requirements.txt

Then: Run the command in the same root folder:

python -m spacy download en

• You can skip this

Tensorflow version:

• For my case I ran 1.14 for Python 3.7

pip install tensorflow

• You can technically run that and just proceed, but I cannot guarantee you won't have any issues when running it. You will get avx2 etc errors.
• However the new versions beyond 1.15 has code breaking errors between keras and tensorflow. Beware of which versions you install.
• You can look through this repo tensorflow windows. Provided by fo40225.
  • then select which tensorflow version you want together with your python version.
  • download and install via pip
• Alternatively, you can also run intel optimised version of tensorflow from here.

Navigate to the \armchairexpert\config folder:

• Create a copy of armchair_expert.example.py and rename it to armchair_expert.py
• Create a copy of config/ml.example.py and rename it to config/ml.py
• Then go back to the root folder \armchairexpert\
• It is preferred to import data first for training before starting the bot.
  • You can use the provided import script in \scripts\import_text_file.py
  • Simply copy the script to the root folder, together with your txt file and run it as: python import_txt.py "<your-data-file-name-here>"
  • It would look like this: python import_text_file.py "CAN YOU HEAR ME.txt"
  • Another option is to let the bot run for a while and learn from the user messages being sent in the servers.
• Every time the bot starts it will train on all new data it acquired since it started up last -The bots sentence structure model is only trained once on initial start-up. To train it with the most recent acquired data, start the bot with the --retrain-structure flag. If you are noticing the bot is not generating sentences which the structure of learned material, this will help.

• Before you can run the bot, you will need to have a Discord app account and register a discord bot to interface with.
• You can take a look and see a step by step instructions on how to create and add the bot into your server, Here.

• Navigate to the \armchairexpert\config folder:
• Create a copy of discord.example.py and rename it to discord.py
• Open the discord.py file with your text editor and fill the required fields.
• Now copy the Client ID, Username#xxxx, and TOKEN to the discord.py file.
• Change the configuration as you wish, for instance enabling learn from all, etc.
• Go back to the \armchairexpert\ Directory
• open console within the directory and run, python armchair.py
• When the bot starts you should see a message print to the console containing a link which will allow you to join the bot to a server.

You can just skip this if you want discord.

• You will need to create an application on the twitter developer site on your bot's twitter account https://apps.twitter.com
• After creating it, assign it permissions to do direct messages (this isn't default)
• Create an access token for your account
• Copy config/twitter.example.py to config/twitter.py
• Fill in the tokens and secrets along with your handle
• Go back to the Root Directory
• python armchair.py