Towards a Joint Khmer Text Recognition and Word Segmentation
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionComputation and Language
AI summaryⓘ
The authors worked on making it easier to read and understand Khmer text from images. Normally, computers first read the text and then figure out where the words are because Khmer doesn't use spaces. Their new approach combines these two steps into one model, which is faster and less likely to make mistakes. Tests showed their method works well on various types of documents and handwriting by recognizing characters and marking where words start and end all at once.
Khmer text recognitionword segmentationconnectionist temporal classification (CTC)document image processingOCRcharacter recognitiondocument modalitieshandwritten text recognitionretrieval-augmented generation (RAG)
Authors
Marry Kong, Rina Buoy, Sovisal Chenda, Nguonly Taing, Masakazu Iwamura, Koichi Kise
Abstract
Text recognition, or extracting electronic text from document images, has been indispensable for knowledge retrieval tasks, such as retrieval-augmented generation (RAG). For Khmer, extracted text is subject to an extra word segmentation step, as Khmer does not use any visible word delimiters to denote word boundaries. Thus, a recognition-then-segmentation pipeline for Khmer requires two separate sequential models; this is not only error-prone but also adds significant latency for large-scale document processing. This paper proposes a novel joint Khmer text recognition and word segmentation framework in a unified model. The proposed model, using a connectionist-temporal-classification (CTC) decoder for fast, parallel decoding, can be instructed to recognize Khmer text with ($b=1$) and without ($b=0$) word segmentation. Experimental results on different benchmark datasets of different document modalities (document, scene, and handwritten images) show that the proposed model can not only recognize characters in document images but also locate word boundaries, removing the need for an extra word segmentation step in a conventional sequential pipeline.