Mentor: Ishika Singh
Team 1
Team Members: Yatish Goel, Varenya Srivatava, Tushar Singla, Nishima Panwar, Gagan Aryan
Abstract:
Twitter according to Wikipedia is a micro-blogging and social networking site where users post and interact with messages known as âtweetsâ. These tweets play an integral part in conveying many important messages. A new policy by the government, the retirement of a sportsman, key updates of tech-giants and many other updates are announced first on twitter. Over a billion tweets are tweeted every day. With all these tweets getting circulated every second, it is hard to tell whether the sentiment behind a specific tweet will impact a company, or a personâs, brand for being viral(positive), or devastate profits because it strikes a negative tone. Hence, capturing the sentiment of a tweet is critical. In this project, we not only try to capture a sentiments tweet but also try to extract the phrase that conveys a particular sentiment. To achieve this we have used two NLP techniques NER and Roberta trained models by modelling the problem statement in the form of a question answering task. We obtained an accuracy of 66.4 per cent using a baseline NER model and an accuracy of 70.3 per cent using the Roberta model. This task was also a Kaggle competition that concluded recently. We finished 1313 in the competition out of 2227 team with an accuracy of 71.25 per cent.
Documentation: Link
Team 2
Team Members: Atul Yaduvanshi, Saad Ahmad, Sajal Goyal, Saurabh K. Gupta
Abstract:
In order to keep cognition models accessible, we need it to understand human language. The Natural Language Processing (NLP) becomes important which is concerned with the interactions between computers and human (natural) languages. One of the basic attributes of human communication is Sentiment. It is important that machines understand these sentiments.
Step 1: Understanding functioning of LM’s and textual data preprocessing.
Step 2: Utilizing pretrained LM’s and building a basic pipeline of the solution.
Step 3: Fine tune LM’s.