An In-depth Analysis of OCR Errors for Unconstrained Vietnamese Handwriting

Nguyen, Quoc-Dung; Le, Duc-Anh; Phan, Nguyet-Minh; Zelinka, Ivan

doi:10.1007/978-3-030-63924-2_26

Cited by 5 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following our study [47] on the VNOnDB-OCR dataset, the errors with edit distances 1 and 2 contribute more than 80% of the total errors whereas the ones with the edit distance 1 account for over 60%. In addition, it is shown in [48] that the edit distances of 1 and 2 obtain better correction results instead of higher edit distances.…”

Section: ) Error Correctionmentioning

confidence: 87%

“…Based on the same VNOnDB-OCR dataset, our study [47] provided the error rates at the syllable level and character level as shown in Table 2. We can see that the error rates in the test dataset (4.7% and 13.6% at the character level and syllable level, respectively) are much higher than those in the combined training and validation dataset (1.7% and 4.3%).…”

Section: A Datasetmentioning

confidence: 99%

“…The ''→'' symbol represents the left GT syllable (or phrase) that is incorrectly recognized as the right OCR syllable (or phrase). For a detailed description of the character-level and syllable-level errors and their possible causes, see our previous study [47]. Our proposed OCR post-processing model will focus on detecting and correcting character-level errors.…”

Section: ) Error Detectionmentioning

confidence: 99%

“…According to our study [47] on the same dataset, the top mappings at both character level and syllable level are of one-or two-character length of edits. Hence, we employ character edits with length of one or two characters to generate correction candidates.…”

Section: ) Error Correctionmentioning

confidence: 99%

See 3 more Smart Citations

An Efficient Unsupervised Approach for OCR Error Correction of Vietnamese OCR Text

et al. 2023

Self Cite

View full text Add to dashboard Cite

Different types of OCR errors often occur in OCR texts due to the low quality of scanned document images or limitations in OCR software. In this paper, we propose a novel unsupervised approach for OCR error correction. Correction candidates for OCR errors are generated and explored in their neighborhoods using correction character edits controlled by an adapted hill-climbing algorithm. Correction characters are extracted from only original ground truth texts, which do not depend on OCR texts in training data. A weighted objective function used to score and rank correction candidates is heuristically tested to find optimal weight combinations. The proposed model is evaluated on an OCR text dataset originating from the Vietnamese handwritten database in the ICFHR 2018 Vietnamese online handwritten text recognition competition. The proposed model is also verified concerning its stability and complexity. The experimental results show that our model achieves competitive performance compared to the other models in the ICFHR 2018 competition.INDEX TERMS OCR, character edit, error correction, attention-based encoder-decoder, hill climbing.

show abstract

Section: ) Error Correctionmentioning

confidence: 87%

Section: A Datasetmentioning

confidence: 99%

Section: ) Error Detectionmentioning

confidence: 99%

Section: ) Error Correctionmentioning

confidence: 99%

See 2 more Smart Citations