Sentence Highlighter with BERT Embeddings

Visualizing sentence similarity using BERT embeddings and cosine similarity.

Built an interactive tool to highlight sentence similarities within documents using deep learning embeddings.

  • Processed and tokenized text from .txt documents using NLTK’s sent_tokenize.
  • Converted sentences into embeddings with bert-base-uncased model from Hugging Face Transformers.
  • Computed pairwise cosine similarities between sentences to detect semantic closeness.
  • Generated a color-coded HTML output where similar sentences are visually highlighted.
  • Fully customizable tool: supports model selection, color themes, and exportable reports.

➡️ View Project on GitHub