Proceedings: GI 2018

Supporting Chinese Character Educational Interfaces with Richer Assessment Feedback through Sketch Recognition

Tianshu Chu (Texas A&M University, USA), Paul Taele (Texas A&M University, USA), Tracy Hammond (Texas A&M University, USA)

Proceedings of Graphics Interface 2018: Toronto, Ontario, 8-11 May 2018, 50 - 57

DOI 10.20380/GI2018.08

  • BibTex

    @inproceedings{Chu:2018:10.20380/GI2018.08,
    author = {Chu, Tianshu and Taele, Paul and Hammond, Tracy},
    title = {Supporting Chinese Character Educational Interfaces with Richer Assessment Feedback through Sketch Recognition},
    booktitle = {Proceedings of Graphics Interface 2018},
    series = {GI 2018},
    year = {2018},
    isbn = {978-0-9947868-3-8},
    location = {Toronto, Ontario},
    pages = {50 -- 57},
    numpages = {8},
    doi = {10.20380/GI2018.08},
    publisher = {Canadian Human-Computer Communications Society / Soci{\'e}t{\'e} canadienne du dialogue humain-machine},
    keywords = {Sketch recognition, Chinese, intelligent tutoring system, language learning, handwriting recognition, writing assessment},
    }

Abstract

Students of Chinese as a Second Language (CSL) with primarily English fluency often struggle with the language's complex character set. Conventional classroom pedagogy and relevant educational applications have focused on providing valuable assessment feedback to address their challenges, but rely on direct instructor observation and provide constrained assessment, respectively. We propose improved sketch recognition techniques to better support Chinese character educational interfaces' realtime assessment of novice CSL students' character writing. Based on successful assessment feedback approaches from existing educational resources, we developed techniques for supporting richer automated assessment, so that students may be better informed of their writing performance outside the classroom. From our evaluations, our techniques achieved recognition rates of 91% and 85% on expert and novice Chinese character handwriting data, respectively, greater than 90% recognition rate on written technique mistakes, and 80.4% f-measure on distinguishing between expert and novice handwriting samples, without sacrificing students' natural writing input of Chinese characters.