Linguistic Features of Discourse within an Algebra Online Discussion Board
Michelle Banawan, Renu Balyan, Jinnie Shin, Walter Leite, Danielle McNamara
Jul 02, 2021 14:10 UTC+2
—
Session PS2
—
Gather Town
Keywords: Computational linguistics, online learning, discussion posts
Abstract:
This study investigates the dimensions of language in students’ discourse within an online learning platform, Math Nation (formerly known as Algebra Nation). Discussion board posts and comments as well as the performance and demographics data in Math Nation were quantitatively examined to establish any existing relationships. The predictive models reveal that linguistic features that estimate cohesion and sentiments that depict social elements (e.g. affect friends and family and social order component) account for 9.7% of the variance of the students’ scores in the Algebra I exam taken at the end of course (EOC). The results did not show any significant variations on the students’ Algebra I scores as a function of gender, grade level or district factors. The most significant linguistic predictors were cohesion indicators that included word-specific attributes, such as length variations, frequency, psycho-linguistic properties (e.g., familiarity, age of acquisition, concreteness, imageability and meaningfulness). It also included features measuring uniformity and consistency of sentence constructions, overlaps (noun and argument), connectives, and lexical diversity. The subsequent principal components analysis performed to profile the online discussion board discourse shows that the texts are distinguished by linguistic dimensions that measure lexical sophistication, syntactical construction, referential cohesion, lexical diversity, sentence construction, semantics, and sentiment. These principal components were discriminative of the discussion board discourse accounting for more than 30% of the variance in the texts. Overall, this study confirms the presence of various dimensions of cohesion and social sentiment in the language used within collaborative learning environments and the important role they play in the academic performance of students in terms of Algebra I scores. Further, this study corroborates findings from prior work on the strong predictive value of the Math baseline scores on the Algebra I scores for the subsequent year.