Missing space between words

[hxt365]

Description of the bug | 错误描述

rapid_table model fails to preserve spaces between words. Please take a look at the column titles in the below example.

Consolidatedsales	First9months		Reportedchange	Currency	Organic
(Dollarsinmillions)	2024	2023	(U.S.GAAP)	effects1)	change*
Airbags, Steering Wheels and Other2)	$5,264	$5,191	1.4%	(1.0)%	2.4%
SeatbeltProductsandOther2)	2,511	2,533	(0.9)%	(1.2)%	0.3%
Total	$7,774	$7,724	0.7%	(1.1)%	1.7%
Americas	$2,637	$2,665	(1.0)%	(0.2)%	(0.8)%
Europe	2,231	2,122	5.2%	1.4%	3.7%
China	1,423	1,488	(4.4)%	(2.0)%	(2.3)%
Asiaexcl.China	1,483	1,449	2.3%	(5.3)%	7.7%
Total	$7,774	$7,724	0.7%	(1.1)%	1.7%

How to reproduce the bug | 如何复现

Run magic-pdf on the provided table.

Operating system | 操作系统

Linux

Python version | Python 版本

3.10

Software version | 软件版本 (magic-pdf --version)

1.0.x

Device mode | 设备模式

cuda

[myhloli]

Currently, the OCR model used for table parsing is a Chinese-English mixed model. For some English words that are close to each other, there may be issues with missing spaces. We do not have plans to fix this issue in the short term. A possible solution is to replace the existing RapidOCR backend with one that offers better performance for English text parsing:

MinerU/magic_pdf/model/sub_modules/table/rapidtable/rapid_table.py

Line 23 in 4d70b16

self.ocr_model_name = "RapidOCR"

Please replace the original RapidOCR with a more suitable alternative as indicated.