Medical Abstract Sequential Sentence Classifier (NLP)

Check out the project here

• The State of the Art Model was Implemented from PubMed 200k

  Tools Used:

  NLP, Analyze Text Data, Tokenization, Embeddings, Text Vectorization, Leveraging Pretrained Sentence Encoder, Implement Model from Research Publication, Naive Bayes, Conv1D, Multimodal Models, Bidirectional LSTM, Text Data Preprocessing, Categorical Encoding, One Hot Encoding, Handling Missing Variables, Model Evaluation on Common Metrics (Accuracy, Precision, Recall, F-1), Hyperparameter Tuning, Optimize Model Performance, Data Manipulation, Pandas, Numpy, Concatenating Multiple Model Heads

  Summary:

• Proficient in NLP techniques: Preprocessed and analyzed text data, including tokenization, embedding, and sequence padding, to prepare it for modeling.

• Experienced in building and evaluating machine learning models: Developed and compared various models using techniques like token embeddings, character embeddings, and multimodal fusion to achieve optimal classification performance.

• Skilled in deep learning frameworks: Utilized TensorFlow and Keras to implement complex neural network architectures, including Conv1D and Bidirectional LSTMs, for text classification tasks. Proficient in data preprocessing and manipulation: Processed and cleaned text data, handled categorical variables using encoding techniques, and computed statistical measures to understand data characteristics.

• Strong analytical and problem-solving skills: Analyzed model predictions, visualized evaluation metrics, and iteratively improved model performance to achieve desired outcomes.