講演抄録/キーワード |
講演名 |
2018-12-13 10:45
An attention-based encoder-decoder for recognizing Japanese historical document recognition ○Le Duc Anh(CODH)・Mochihashi daichi(ISM)・Masuda katsuya・Mima Hideki(UT) PRMU2018-78 |
抄録 |
(和) |
attention model, encoder-decoder approach |
(英) |
Inspired by the recent successes of attention based encoder-decoder (AED) approach on image captioning, machine translation, we present an AED model as an end-to-end recognition system for recognizing Japanese historical document. The recognition system has two main modules: a dense convolution neural network for extracting multiscale features, and a Long Shor Term Memory (LSTM) decoder with attention model for generating target text. We can train the model end-to-end. The model requires only input text line images and corresponding output characters. Therefore, we don’t need the annotation in character level and save a lot of time for making annotations. The recognition system is trained by our annotated documents. We show the data imbalance problem in the current data and its effect on the performance of the recognition system through the experiments. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Japanese historical document / attention model / encoder-decoder approach / / / / / |
文献情報 |
信学技報, vol. 118, no. 362, PRMU2018-78, pp. 19-22, 2018年12月. |
資料番号 |
PRMU2018-78 |
発行日 |
2018-12-06 (PRMU) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
PRMU2018-78 |