🚧 WORK IN PROGRESS - NLPKernels

🚩 A collection of NLP Challenge Kernels competed on Kaggle

Attending to the Disaster using Transformers

Overview

1. Data Preprocessing
   1. Remove Unwanted data from the tweets including urls, user names incorrect words
   2. Remove Stopwords and tokenize using HuggingFace BertWordPiece Tokenizer
2. Exploratory Analysis
   1. What a disaster tweet looks like ?
   2. Plot Disaster Tweets most common words by category
3. Classify Tweets
   1. Use simple pre-trained glove embeddings
   2. Try BERT (using HuggingFace Implementation)
   3. Try Roberta (using HuggingFace Implementation)
   4. Tweet Classification models results summary