Predication of Writing Originality Based on Computational Linguistics

Area of Science:

• Computational Linguistics
• Educational Assessment
• Natural Language Processing

Background:

• Traditional writing originality assessments rely heavily on subjective scoring, limiting reliability and objectivity.
• There is a need for more objective and quantifiable methods to evaluate writing originality.
• Computational approaches offer potential for analyzing linguistic features and semantic structures.

Purpose of the Study:

• To investigate the efficacy of topic analysis and semantic networks in assessing writing originality.
• To develop and validate computational linguistic features for quantifying writing originality.
• To explore the influence of reference groups on originality scoring.

Main Methods:

• Latent topic modeling was used to group essays by topic, creating refined reference groups.
• Semantic network features (path distance, semantic differences, centrality, similarity) were extracted from student essays.
• These features were analyzed to predict writing originality scores.

Main Results:

• Writing originality is influenced by both intrinsic text characteristics and the evaluated reference group.
• Computational linguistic features, including semantic network metrics, significantly predict writing originality in Chinese essays.
• Each of the four feature categories demonstrated predictive power, though effectiveness varied by topic.

Conclusions:

• Computational linguistic analysis provides an objective and effective method for assessing writing originality.
• The study validates the use of topic analysis and semantic networks in educational assessment.
• Feature analysis offers valuable insights to support human raters in originality scoring.