The way individuals/People are expressing their feelings, thoughts, and opinions has entirely transformed in the
previous few years thanks to the discovery of social sites, websites, blogs, wikis, and other online cooperative media
and web sources. The purification of data from a large amount of unstructured data on the Web can be a key feature for
dealers/sellers who want to develop a mental model for customers of their product or brand. These operational social
data, however, remain hardly reachable to machines, as they are exactly meant for human consumption.
Proposed Methodology
We are supposed to propose a system, that can detect if the given tweet is positive or negative, and how many people are in favor or opposition to that tweet(Topics like Machine Learning, Deep Learning, and so on).
We are using Python for data preprocessing, we’re using social networking services scrape which is a Python-based package used for fetching real-time tweets, and for front-end and Python web integration we will be using Django.
Dataset And Model Discussion
We have used the dataset named Sentiment140 [2]. Which contains 1.6M tweets labeled data. It is labeled with two classes Positive and negative. Which meets our requirements to use this dataset for training the model.
We used Roberta pre-trained as well as created our own model to analyze the tweets. Obviously, at this level, Roberta is performing much better than our model but we trying to make our model as accurate and efficient as possible.
How to Project?
Clone the repository
Download models zip file
and extract
in the project folder.
Run the pip install -r requirements.txt to install all packages in the project folder.
