Essay + prompt generation

Read the dissertation

My Masters dissertation focused on investigating natural language generation techniques for open-ended prompts and free-text responses for language learning and assessment.

At the time of submission, no work in the education domain had focused on the generation of open-ended questions and free-form essay responses. Potential applications of this work include:

• generating synthetic essays at particular levels to create balanced datasets for other educational NLP tasks
• providing exemplar work to students at reduced cost
• diversifying exam question prompts at reduced cost

This work contrasted four language generation techniques alongside a KNN baseline. All models were trained using open-ended prompts and essays from the Cambridge Learner Corpus. The models were also evaluated for ability to generalise to other examinations on the International Corpus of Learner English (ICLE).

This work required use of Python and NLP libraries and models such as Marian-NMT and OpenAI GPT-2.