Encoder-decoder models have made great progress on handwritten mathematical expression recognition recently. However, it is still a challenge for existing methods to assign attention to image features accurately. Moreover, those encoder-decoder models usually adopt RNN-based models in their decoder part, which makes them inefficient in processing long $\LaTeX{}$ sequences. In this paper, a transformer-based decoder is employed to replace RNN-based ones, which makes the whole model architecture very concise. Furthermore, a novel training strategy is introduced to fully exploit the potential of the transformer in bidirectional language modeling. Compared to several methods that do not use data augmentation, experiments demonstrate that our model improves the ExpRate of current state-of-the-art methods on CROHME 2014 by 2.23%. Similarly, on CROHME 2016 and CROHME 2019, we improve the ExpRate by 1.92% and 2.28% respectively.
翻译:编码器- 解码器模型最近在手写数学表达式识别方面取得了巨大进展。 然而, 对现有方法来说, 准确关注图像特征仍是一项挑战。 此外, 这些编码器- 解码器模型通常在其解码器部分采用基于 RNN 的模型, 这使得它们在处理长$\ LaTeX ⁇ ⁇ $序列方面效率低下。 在本文中, 使用一个基于变压器的解码器取代基于 RNN 的解码器, 使整个模型结构非常简洁。 此外, 引入了一个新颖的培训战略, 以充分利用变压器在双向语言建模中的潜力。 与一些不使用数据增强功能的方法相比, 实验表明我们的模型将2014年CROHME当前最新工艺方法的开发率提高了2. 23%。 同样, 在 CROHME 2016 和 CROHME 2019 上, 我们分别提高了1.92 % 和 2.28% 。