Some challenges in automatic English text correction
Tatiana Al-Chueyr Martins | Sunday 10:15 | Room D
Some applications in the market assist users to correct different writing mistakes, including spelling and grammar errors. However, very rarely these tools are used by school teachers. For most of them, it is still time consuming and tedious to correct (beginners) student essays.
This talk will introduce some challenges in automatic English text correction. It will also present how it is possible to use Python libraries (scikit-learn, SciPy and NumPy) in order to spot English mistakes such as: articles, capitalization and spelling.
In order to train and test the classifier, an open dataset will be used: EF-Cambridge Open Language Dataset (https://corpus.mml.cam.ac.uk/efcamdat/).
During the presentation, the accuracy of the implementation will be compared to at least one commercial application and a similar open source tool. It will be discussed how this kind of work can bring value to existing educational applications. Limitations and further steps will also be discussed.
This presentation is a successive work from the cooperation of Education First (language teaching institution) engineers and University of Cambridge language researchers.