Ontology of Academic Sentence Dependencies for a Verb Choice Suggestion

摘要

Non-native researchers encountered a problem of lacking academic writing skill. During the writing, we may accidentally use a word repeatedly due to our familiarity that reduces a quality of writing. To solve the problem, a paraphrase is a good option. It helps the manuscript read more flowery by reducing duplicate words and refining sentence alignments. In this study, we propose a novel idea to use a sentence dependency ontology to suggest possible verbs replaceable on existing context without influence to the original intention. We created an ontology-based system and designed a new ontology. To discover a list of verb choices, our idea is based on sentence dependency, especially a dependency between subject and verb (nsubj) as well as a relationship between verb and object (dobj). We chose them because these two dependencies had a strong relationship to the verb of the sentence. To evaluate the system, we compared the efficiencies of two different systems, i.e., a tradition system utilizing synonyms as word choices and our ontology-based system. As the results, ours provided the better performance rather than the traditional system. This clarifies that our system should be a proper solution for studies on paraphrasing.