pip install nltk
pip install numpy
pip install keras
pip install tensorflow
pip install spacy
download spacy model
python -m spacy download en_core_web_sm
When first running the program, it may ask you to download certain NTLK packages. The only way I found I could download them was by running the code below:
import nltk
import ssl
try:
_create_unverified_https_context = ssl._create_unverified_context
except AttributeError:
pass
else:
ssl._create_default_https_context = _create_unverified_https_context
nltk.download()Link to source can be found here: https://stackoverflow.com/questions/38916452/nltk-download-ssl-certificate-verify-failed
pip install googletrans
pip install wikipediaapi
To train the chatbot you can run the file app/chatbot/train.py
To use the chatbot you can run the file app/chatbot/chatbot_app.py
The bot has been updated to handle topics that are out of its range, and respond in a way that directs the conversation back to known conversation areas.
Several conversation structures which indicate symptoms of schizophrenia have been added to the bots repertoire, in particular hallucinations and general disorganization. There is also referal to proper health resources if the bot detects these symptoms.
The POS (Parts of Speech) tagger function generates a tag (noun, verb, adverb, adjective) for every word that gets passed to it. The tag gets passed to the lammetizer function, along with the word, which will then lammetize the word appropriately. Previously, the lammetizer would default every word as a noun. For example, our chatbot will now be able to identify between an "accident" (noun) and "accidentally" (adverb).
The Synonym Recognation function uses WordNet (collection of words and vocabulary) to find synonyms of the words used in our json file. These synonyms get added to the words.pk file which are later used to find the similar words of the training data. For example, the chatbot will now understand sad as also pitiful or distressing and answer appropriately.
Named Entity Recognition is implemented through spaCy (it can also be installed using pip for python). It allows the program to differentiate between different named objects, such as Organizations, People, Geo-Political Entities, Quantities, and many more. The chatbot will mainly make use of this by identifying organizations when asking about work habits. The chabot is able to use the context of the sentence and pretrained knowledge to determine that 'K2' and the 'UN' are both organizations, and favour dialogues responding to work:
