Transformer-based encoder-decoder networks have recently achieved impressive results in handwritten text recognition, partly thanks to their auto-regressive decoder which implicitly learns a language model. However, such networks suffer from a large performance drop when evaluated on a target corpus whose language distribution is shifted from the source text seen during training. To retain recognition accuracy despite this language shift, we propose an external n-gram injection (NGI) for dynamic adaptation of the network's language modeling at inference time. Our method allows switching to an n-gram language model estimated on a corpus close to the target distribution, therefore mitigating bias without any extra training on target image-text pairs. We opt for an early injection of the n-gram into the transformer decoder so that the network learns to fully leverage text-only data at the low additional cost of n-gram inference. Experiments on three handwritten datasets demonstrate that the proposed NGI significantly reduces the performance gap between source and target corpora.

Transcription of handwritten words in historical documents is still a difficult task. When processing huge amount of pages, document-centered approaches are limited by the trade-off between automatic recognition errors and the tedious aspect of human user annotation work. In this article, we investigate the use of inter page dependencies to overcome those limitations. For this, we propose a new architecture that allows the exploitation of handwritten word redundancies over pages by considering documents from a higher point of view, namely the collection level. The experiments we conducted on handwritten word transcription show promising results in terms of recognition error and human user work reductions.