{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/pattern-generation-strategies-for-improving","title":"Pattern Generation Strategies for Improving Recognition of Handwritten Mathematical Expressions","arxiv_id":"1901.06763","date":"2019-01-21","proceeding":null,"authors":["Anh Duc Le","Bipin Indurkhya","Masaki Nakagawa"],"abstract":"Recognition of Handwritten Mathematical Expressions (HMEs) is a challenging\nproblem because of the ambiguity and complexity of two-dimensional handwriting.\nMoreover, the lack of large training data is a serious issue, especially for\nacademic recognition systems. In this paper, we propose pattern generation\nstrategies that generate shape and structural variations to improve the\nperformance of recognition systems based on a small training set. For data\ngeneration, we employ the public databases: CROHME 2014 and 2016 of online\nHMEs. The first strategy employs local and global distortions to generate shape\nvariations. The second strategy decomposes an online HME into sub-online HMEs\nto get more structural variations. The hybrid strategy combines both these\nstrategies to maximize shape and structural variations. The generated online\nHMEs are converted to images for offline HME recognition. We tested our\nstrategies in an end-to-end recognition system constructed from a recent deep\nlearning model: Convolutional Neural Network and attention-based\nencoder-decoder. The results of experiments on the CROHME 2014 and 2016\ndatabases demonstrate the superiority and effectiveness of our strategies: our\nhybrid strategy achieved classification rates of 48.78% and 45.60%,\nrespectively, on these databases. These results are competitive compared to\nothers reported in recent literature. Our generated datasets are openly\navailable for research community and constitute a useful resource for the HME\nrecognition research in future.","url_abs":"http://arxiv.org/abs/1901.06763v1","url_pdf":"http://arxiv.org/pdf/1901.06763v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"pattern-generation-strategies-for-improving","repo_url":"https://github.com/ducanh841988/Artificial-Online-Handwritten-Mathematical-Expressions","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1901.06763","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}