📄 Facilitating English Grammar Learning by a Personalized Mobile-Assisted System With a Self-Regulated Learning Mechanism

Summary

Mobile language-learning research has mostly focused on vocabulary, listening and speaking, while grammar has been largely overlooked. Learners studying alone on phones also need strong self-regulation to decide what to study, track their progress and keep going when the material gets hard. The authors built a personalized mobile system for English-as-a-foreign-language (EFL) grammar that supports self-regulated learning (SRL) through three functions: exercise recommendations, feedback, and an e-portfolio. They tested it in a semester-long quasi-experiment. One group of students used every function to study grammar and do exercises. A control group used the same system only to hand in weekly assignments. The full-access group scored significantly higher on grammar tests, and male and female students benefited to a similar degree. The authors argue that the system works best as an add-on to classroom teaching, because it lets learners plan and monitor their own study without a teacher present. They also see the study as support for building SRL scaffolding into mobile learning tools.

Key points

• Quasi-experimental design with an experimental group of n = 278 and a control group of n = 320, run over one semester.
• The SRL mechanism consists of recommendations, feedback and an e-portfolio package. Recommendations take into account the learner's goals, current level and earlier performance.
• Both groups used the same platform. The control condition was limited to weekly assignment submission, so the comparison isolates the SRL and personalization features.
• The experimental group scored significantly higher on English grammar tests. Gender made no significant difference to the gain, which matches Chen et al. (2019) but differs from Chen and Huang (2014).
• Related work cited includes Grammar Clinic (Li and Hegelheimer, 2013), where learners find and fix sentence-level errors such as fragments and run-ons, and a mobile grammar game with concept mapping (Chu et al., 2017).

Interesting ideas

• I1. The system chooses the next exercise from the learner's goal, current level and past results, not from a fixed sequence, so practice goes where the learner is weakest.
• I2. The control group used the same tool with its key functions switched off, which separates the effect of the mechanism from the effect of simply using the platform.
• I3. The e-portfolio keeps a lasting record of the learner's attempts and progress, which lets the learner monitor themselves instead of relying on a teacher to judge each step.
• I4. Beyond the overall effect, the authors checked whether gains were spread evenly across a subgroup (gender) to catch hidden unequal benefit.
• I5. Grammar Clinic, a tool from the related work, trains learners to spot structural sentence errors such as fragments and run-on sentences.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (learn): Training and self-play could pick the next batch by per-construction error history, putting more effort into the UD relations or frame types where the parser or reader still fails most.
• I2 (learn): Evaluations could run the same pipeline with a single component disabled (for example, repair rules or teacher tiers) as a built-in negative control, so measured gains are attributed to the right part.
• I3 (answer): Student readers could keep a portfolio of failed school-test and word-problem items with the step that broke, giving a clear trail for later diagnosis and retesting.
• I4 (learn): Scores could be routinely split by text genre or sentence type to show whether an improvement helps fables, maths problems and textbook prose alike or only one slice.
• I5 (tree): A UD tree with no finite root, or with several unconnected main clauses, could be flagged as a likely fragment or run-on, which would help when cleaning noisy input or checking generated text.