Improving Text Recognition using Optical and Language Model Writer Adaptation

Abstract

State-of-the-art methods for handwriting text recognition are based on deep learning approaches and language modeling that require large data sets during training. In practice, there are some applications where the system processes mono-writer documents, and would thus benefit from being trained on examples from that writer. However, this is not common to have numerous examples coming from just one writer. In this paper, we propose an approach to adapt both the optical model and the language model to a particular writer, from a generic system trained on large data sets with a variety of examples. We show the benefits of the optical and language model writer adaptation. Our approach reaches competitive results on the READ 2018 data set, which is dedicated to model adaptation to particular writers.

Boosting Offline Handwritten Text Recognition in Historical Documents With Few Labeled Lines

“…In the HTR problem with a reduced training set, TL was applied by Soullard et al in [7]. The main idea behind TL is initializing the parameters of a model by those learned from a huge dataset beforehand, denoted as source.…”

“…Hence, with TL we start learning a different task to avoid learning the whole set of parameters from scratch, preventing overfitting and favoring convergence. In [7], they proposed a method that applies TL in both the optical and the language model. In this and other similar previous proposals on TL, the authors applied DA in both training and test steps.…”

Learning to Read L’Infinito: Handwritten Text Recognition with Synthetic Training Data

“…For big historical datasets, in [5], the authors demonstrated the benefits of carefully designed data augmentation. Another strategy is to apply transfer learning [12,13,25,15,1], i.e., pretraining the HTR model on a big HTR dataset and fine-tuning it on the small training set of the dataset of interest. For HTR on small single-writer historical datasets, pretraining plus fine-tuning has been proven to be a more effective strategy than data augmentation [1].…”

Boosting offline handwritten text recognition in historical documents with few labeled lines

“…In the HTR problem with a reduced training set, TL was applied by Soullard et al in [7]. The main idea behind TL is initializing the parameters of a model by those learned from a huge dataset, denoted by source.…”

“…Also, classification and indexing of transcript text can be easily automated. Handwritten text recognition (HTR) tasks in historical datasets have been faced by many authors in the last few years [1][2][3][4][5][6][7][8][9][10]. In HTR, transcribing each author can be considered a different task, since the distribution of both model input and output varies from writer to writer.…”