digital restoration for printed text manuscripts

  • Unique Paper ID: 162679
  • Volume: 10
  • Issue: 10
  • PageNo: 1083-1086
  • Abstract:
  • Text recovery is a complex problem in natural language processing, involving the recovery of partially destroyed text. Deep learning techniques, like the Seq2Seq model, have demonstrated significant promise for text recovery tasks in the past couple of years. In this paper, we suggest the GloVe integrated Seq2Seq model for text recovery. This approach aims to exploit the semantic information contained in the GloVe embedding to translate partially destroyed text. In order to assess how successful the suggested approach is, on carrying out a sensitivity analysis to investigate how various hyperparameters affect the model's functionality. The analysis shows that the choice of hyperparameters, such as hidden layer size and learning rate, can significantly affect model performance, also by perform preprocessing steps such as data cleaning and augmentation to improve input data quality. GloVe embedding, which encodes semantic information of words and sentences in a dense vector space, is the primary strength of the suggested approach. This enables the model to interpret the input text's meaning even when it is entirely or partially distorted. The model's accuracy is further increased by using attention techniques, which enable the model to concentrate on pertinent segments of the input sequence.

Cite This Article

  • ISSN: 2349-6002
  • Volume: 10
  • Issue: 10
  • PageNo: 1083-1086

digital restoration for printed text manuscripts

Related Articles