Converting Image Files to LaTeX Format Using Computer Vision, Natural Language Processing, and Machine Learning

Authors : Murat Kazanç, Tolga Ensari, Mustafa Dağtekin

Pages : 253-266

Doi:10.26650/acin.1258719

View : 56 | Download : 89

Publication Date : 2023-12-29

Article Type : Research Paper

Abstract :A few decades ago, people used printed resources such as books and magazines to learn. With the development of technology, digital documents have replaced printed resources. These documents can occur in the form of images or various text formats. Many different applications exist for preparing digital documents, one of these being LaTeX. LaTeX is a document preparation system and typesetting software that is used especially in the field of scientific publications and mathematics for preparing high quality documents. When preparing a document using LaTeX, the content is made ready using a markup language, which creates difficulties for some users. However, one of the main advantages of using the LaTeX system is that it distinguishes the document’s content from its formatting. Once the content is created, the formatting can be easily replaced. Generating LaTeX code from an image-formatted document requires both the use of computer vision and NLP. This study discovers the boundaries (blocks) of the places where text, tables, and figures are located on an image before making a text classification using the natural language processing methods of these blocks. The next stage of the study determines the reading order to enable meaningful flow. The final stage of the study produces a LaTeX code using the obtained information.
Keywords : Bilgisayarlı görü, metin sınıflama, okuma sırası, makine öğrenmesi