Search Results (2)

OCR for low-resource languages is still held back by the same small number of issues: too little labeled image-text data, too few benchmarks, and thin language-specific tooling. Kurdish Sorani is a particularly awkward case. It is written in a modified Arabic script, runs right to left, and has orthographic habits that standard Arabic OCR engines handle poorly. This paper describes a transformer OCR system for Sorani trained almost entirely on synthetic data, meaning line images rendered on the fly from a text corpus rather than manually transcribed scans. The pipeline has three parts: corpus-driven line synthesis, a deterministic script-aware normalization step based on character-level transliteration, and a TrOCR encoder–decoder recognizer. Text lines are rendered with randomly sampled fonts and sizes, then passed through stochastic augmentation to mimic realistic distortions. The system is evaluated twice. On an in-distribution synthetic set of 200 rendered lines, the best model reaches a character error rate of 0.0434, a word error rate of 0.1246, and 64.0% exact matches. More importantly, on a real-world test set of 19 scanned Kurdish documents (467 lines, 28,468 characters) processed end-to-end through detection and recognition, it reaches a character error rate of 0.0305 and a word error rate of 0.1770, beating both Arabic and Kurdish Tesseract baselines and an existing Kurdish TrOCR model while being considerably smaller than the latter. A controlled ablation, in which eight variants are trained under one shared budget and scored on identical images, then isolates what each design choice contributes. The label space is the largest design effect, and the reason is concrete: the decoder’s pre-trained tokenizer has no representation for seven common Sorani graphemes, which cover 14.7% of the corpus and place a floor under any model trained on native-script labels. Corpus size dominates overall and behaves as a threshold, font diversity helps with diminishing returns, and stochastic augmentation buys robustness at a small cost in in-distribution accuracy. Aligning detected lines against the transcribed ones further shows that line detection contributes under 1% of the reported character error on this material. The broader point, at least for Sorani, is that the synthetic training data and the label space in which the model predicts have to be designed together: a compact recognizer built that way outperforms a substantially larger released Kurdish model on genuine document images. Full article

Applications based on Long-Short-Term Memory (LSTM) require large amounts of data for their training. Tesseract LSTM is a popular Optical Character Recognition (OCR) engine that has been trained and used in various languages. However, its training becomes obstructed when the target language is not resourceful. This research suggests a remedy for the problem of scant data in training Tesseract LSTM for a new language by exploiting a training dataset for a language with a similar script. The target of the experiment is Kurdish. It is a multi-dialect language and is considered less-resourced. We choose Sorani, one of the Kurdish dialects, that is mostly written in Persian-Arabic script. We train Tesseract using an Arabic dataset, and then we use a considerably small amount of texts in Persian-Arabic to train the engine to recognize Sorani texts. Our dataset is based on a series of court case documents in the Kurdistan Region of Iraq. We also fine-tune the engine using 10 Unikurd fonts. We use Lstmeval and Ocreval to evaluate the outputs. The result indicates the achievement of 95.45% accuracy. We also test the engine using texts outside the context of court cases. The accuracy of the system remains close to what was found earlier indicating that the script similarity could be used to overcome the lack of large-scale data. Full article