LV-ROVER: Lexicon Verified Recognizer Output Voting Error Reduction
This work addresses the problem of improving accuracy in handwritten text recognition for applications like document digitization, though it appears incremental as it builds on the existing ROVER framework.
The paper tackles the challenge of offline handwritten text line recognition by proposing LV-ROVER, a method that reduces complexity and combines hundreds of recognizers without language models, achieving state-of-the-art performance on the RIMES dataset.
Offline handwritten text line recognition is a hard task that requires both an efficient optical character recognizer and language model. Handwriting recognition state of the art methods are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN) coupled with the use of linguistic knowledge. Most of the proposed approaches in the literature focus on improving one of the two components and use constraint, dedicated to a database lexicon. However, state of the art performance is achieved by combining multiple optical models, and possibly multiple language models with the Recognizer Output Voting Error Reduction (ROVER) framework. Though handwritten line recognition with ROVER has been implemented by combining only few recognizers because training multiple complete recognizers is hard. In this paper we propose a Lexicon Verified ROVER: LV-ROVER, that has a reduce complexity compare to the original one and that can combine hundreds of recognizers without language models. We achieve state of the art for handwritten line text on the RIMES dataset.