Sentence Similarity Measures for Fine-Grained Estimation of Topical Relevance in Learner Essays
This work addresses the specific problem of fine-grained topical relevance estimation in learner essays, which is incremental as it builds on existing methods for a niche domain.
The paper tackled the problem of assessing sentence-level prompt relevance in learner essays by evaluating various systems and proposing a new method that adjusts weights of pre-trained word embeddings, achieving substantially higher accuracy compared to baselines.
We investigate the task of assessing sentence-level prompt relevance in learner essays. Various systems using word overlap, neural embeddings and neural compositional models are evaluated on two datasets of learner writing. We propose a new method for sentence-level similarity calculation, which learns to adjust the weights of pre-trained word embeddings for a specific task, achieving substantially higher accuracy compared to other relevant baselines.