Consumer-Complaint-NLP

Installation: Clone the git repo (https://github.com/Samaale11/Consumer-Complaint-NLP.git). Dependency installation: Use pip install command to to install dependencies in the environment.yml file. Data: download consumercomplaints.csv from Kaggle (https://www.kaggle.com/datasets/selener/consumer-complaint-database) and store it in the repository root. Instructions: Open the NLP.ipynb in Jupyter Notebook. Cells have be run used consecutively to load data, preprocess (lowercase, remove punctuation/stopwords, lemmatize), vectorize (CountVectorizer/Tfidfvectorizer), model topics (LDA/NMF) and assess the coherency index (Gensim Cv). Outputs: The outputs include processeddata.csv (complaints processed) and topics ouput based on coherence scores. System requiremets: 8GB of RAM and around 30 minutes per 10k samples, Python 3.8. Interpretation: View topic words and the coherence scores