sankalpjain99/Automatic-Essay-Scoring

The system uses Word2Vec embeddings fed through stacked LSTM layers (300→64 units) with dropout regularization to capture sequential semantic patterns in essays, trained with MSE loss and RMSprop optimization. Beyond initial ML baselines (linear regression, SVR, Random Forest), the architecture incorporates linguistic features—sentence/word counts, POS tag distributions, and misspelling detection—extracted during preprocessing with CountVectorizer for feature engineering. The Flask-based API accepts essays via JSON POST requests and returns predicted scores, making the trained model accessible through a web interface with real-time inference capabilities.

No commits in the last 6 months.