aashishpiitk/tweet-sentiment-extraction

TEAM DOCUMENTATION FOR TWEET SENTIMENT EXTRACTION

PROBLEM STATEMENT

given the tweet the task is to select a part of the tweet which conveys the sentiment of the tweet as neutral, positive or negative.

PLAN OF ACTION

    1. Explore basic of NLP
    2. Explore the dataset and analyse the problem statement
    3. Explore and understand the public notebooks on kaggle and then select the model which suits the team well
    4. Make some changes to the model and then evaluate the model with those changes
    5. Try other variations of the language model 
    6. Understand the innner workings so that suitable  modifications can be made to the head of the model to increase our score

WORK/STUDY DONE TILL NOW

Steps 1 to 4 in above course of action have been completed. Currently we are on the 5th step.

Papers read till now

• roBerta
• word2vec

FUTURE PLAN IN NEXT 10 DAYS

Exploring other variations of BERT  and other language models such as XLNet and trying to apply them to the current problem statement

Papers to read in future

• BERT
• TRANSFORMERS

Contributers

• AASHISH PATEL