Graph-Based Comma Splice Correction Using Sentence Dependency Parsing

Abstract: Grammatical Error Correction (GEC) systems aim to automatically fix errors in text written by language learners. Most current GEC tools focus on word-level and phrase-level mistakes, such as spelling, preposition, and article errors. However, sentence-level errors are also common in learner writing. In this paper, we focus on correcting a specific sentence-level error: the comma splice. Our approach uses sentence dependency parses to detect and correct these errors. On our evaluation dataset, our method achieved 53.70% precision and 28.43% recall.