marufzubery / red-list-bot

Red List bot that informs you about endangered species

Red-List-Bot

Red List bot that informs you about endangered species

How to run chatbot

1. Clone the repository from github https://github.com/samiraliyev573/Red-List-Bot

2. Have python installed (python 3.6+ ) https://www.python.org/downloads/

3. Have anaconda installed. (In windows when downloading anaconda, check the box that says add anaconda to path environment variables.) Install anaconda here: https://docs.conda.io/projects/conda/en/latest/user-guide/install/

4. Now we create anaconda environment so whatever we do will be outside our system and wont do any harm to our currently installed packages. Type 'conda create -n projenv python=3.8' (if this command doesnt work, update your conda by typing 'conda update conda')

5. Now open terminal and navigate to the src folder of the project(You can also do it by opening in visual studio code and selecting new terminal.)

6. Inside the src folder, type 'conda activate projenv'. Now our virtual environment is active inside this folder. Only thing left is to install necessary packages. Run: 'conda install pytorch torchvision torchaudio cpuonly -c pytorch' if in windows or conda install pytorch torchvision torchaudio -c pytorch if in macos. Run 'conda install nltk'

7. Last step is to type 'python chat.py' which will run the python file.

Enjoy the ChatBot ;)

About the ChatBot

This chat bot uses nltk packages and uses Natural Language Processing. All of the code necessary to run it can be found in src folder. Source folder has 6 files.

• nltkproperties.py file basically designed to tokenize, stem and put them in bag of words. Tokenizing will put the string into the words, punctiation and numbers. Example: "How are you will" be changed into "How", "are", "you". Stemming will seperate roots from the actual word. Example: universe and university will be turned into univers. bag_of_words method will create array that is 0 if the sentence has specific word in it or make it 1 if it has that specific work. Purpose of this is to generate aray of 1s and 0s then use it for training our model.
• model.py file is about creating Neural Network for out data. It will create three linear layers, which are hidden layer, input layer and the output layer.
• train.py class is for training the model. Note: If github clone you have downloaded already has data.pth file, you may straight away run chat.py file. It will load contents of intents.json into the intents var. From there, it will use list of which words to ignore and then use stemming and tokenization methods from nlty-utils class. After that, it will sort them all uniquely into the set and start the training model. Once the model is done, it will save it in data.pth so chat.py can use it for chatting with the user.
• chat.py file will read input from the user, tokenize it and then put it in bag of words. Then it will predict the answer. If the probability of the predicted answer is greater than 0.75, it will randomly pick responses from intents.json and present it to the user. If the probability is not high enough, it will simply assume that he has no answer for that question and print I do not understand.
• Intents.json is about tags, patterns and responses. Tags are overall intent and each have pattern and response. Pattern is basically question and response is basically an answer. From the question user asks, it RedListBot will predict appropriate tag and patterns and try to find the response from this file.