Update README.md
Browse filesAutomatic Sentiment Labeling of IMDb Movie Reviews Using Zero-Shot Classification
In this experiment, we aimed to automatically label IMDb movie reviews with sentiment labels (positive or
negative) using a zero-shot classification approach. We utilized the IMDb movie reviews dataset available
through the NLTK library, which contains a collection of movie reviews categorized by sentiment. We
sampled a subset of the dataset for demonstration purposes.
To perform sentiment labeling, we employed the BART-large-mnli model from the Hugging Face
Transformers library. This model is pre-trained on various natural language understanding tasks, including
sentiment analysis, and is capable of generating text classifications without the need for fine-tuning on
specific datasets.
We developed a pipeline that processes each movie review through the BART model, generating a sentiment
label (either positive or negative) based on the content of the review. This process is achieved using zero-shot
classification, where the model predicts the sentiment label without explicit training on sentiment-labeled
data.
The experiment concludes with the creation of a new dataset containing the original IMDb movie reviews
along with the predicted sentiment labels. This dataset provides valuable insights into the sentiment
distribution of IMDb movie reviews and serves as a foundation for further analysis and model training tasks.
Overall, this experiment demonstrates the effectiveness of zero-shot classification models in automatically
labeling text data with sentiment information, enabling efficient sentiment analysis tasks without the need
for extensive manual labeling or model fine-tuning.