Cross-document word matching for segmentation and retrieval of Ottoman divans

Abstract

Cataloged from PDF version of article.Motivated by the need for the automatic\ud indexing and analysis of huge number of documents in\ud Ottoman divan poetry, and for discovering new knowledge\ud to preserve and make alive this heritage, in this study we\ud propose a novel method for segmenting and retrieving\ud words in Ottoman divans. Documents in Ottoman are dif-\ud ficult to segment into words without a prior knowledge of\ud the word. In this study, using the idea that divans have\ud multiple copies (versions) by different writers in different\ud writing styles, and word segmentation in some of those\ud versions may be relatively easier to achieve than in other\ud versions, segmentation of the versions (which are difficult,\ud if not impossible, with traditional techniques) is performed\ud using information carried from the simpler version. One\ud version of a document is used as the source dataset and the\ud other version of the same document is used as the target\ud dataset. Words in the source dataset are automatically\ud extracted and used as queries to be spotted in the target\ud dataset for detecting word boundaries. We present the idea\ud of cross-document word matching for a novel task of\ud segmenting historical documents into words. We propose a\ud matching scheme based on possible combinations of\ud sequence of sub-words. We improve the performance of\ud simple features through considering the words in a context.\ud The method is applied on two versions of Layla and\ud Majnun divan by Fuzuli. The results show that, the proposed\ud word-matching-based segmentation method is\ud promising in finding the word boundaries and in retrieving\ud the words across documents

A deep learning model for Ottoman OCR

“…This project was used by scholars working with the Ottoman archives. As an extension of the previous work in Reference 14, a method was proposed for segmenting and retrieving words in Ottoman divans (books of poetry), motivated by the need for the automatic indexing and analysis of Ottoman divan poetry in Reference 15. Since most diwans have multiple copies with minor differences, a cross document word matching approach was employed for querying and retrieving the text from documents.…”

Printed Ottoman Text Recognition Using Synthetic Data and Data Augmentation

“…There is quite a limited number of studies dealing with Ottoman documents in the literature. Most of them are focusing on retrieval of Ottoman documents using traditional computer vision techniques used to map word shapes [10,11,13,18,56]. Some others constitute the early works on Ottoman character recognition using conventional machine learning algorithms with quite limited proprietary datasets [32,33,42,67].…”

Transcription of Ottoman Machine-Print Documents

“…A very limited number of studies on text recognition in Ottoman Turkish have been identified in the literature. Most of them are dated to pre-deep learning era and use traditional machine learning techniques [12,17,18,26]. In a study that used deep learning techniques on Ottoman documents for the first time, Aydemir et al trained an RNN system by manually extracting features from a dataset containing 169,148 discrete handwritten word images obtained from population registration documents [13].…”

“…Actually, Ottoman document recognition is a problem that has been attempted for many years, without a sufficiently successful solution. Most of the previous works are on document retrieval tasks using traditional machine learning methods [12,17,26]. Their modest success rates can be attributed to limited sizes of the datasets they use.…”