Using machine learning to predict children’s reading comprehension from linguistic features extracted from speech and writing.

Abstract

Advances in machine learning (ML) are poised to contribute to our understanding of the linguistic processes associated with successful reading comprehension, which is a critical aspect of children's educational success. We used ML techniques to investigate and compare associations between children's reading comprehension and 260 linguistic features extracted from their speech and writing. Language samples were gathered from 99 linguistically diverse children in grades 4-6 using Talk2Me, Jr., an online language and literacy assessment platform. Lexical and syntactic features were extracted via a consolidated natural language processing pipeline. We compared five machine learning models predicting reading comprehension from the linguistic features and then, using the best models, analyzed the 20 top predictive features for both the oral-elicited and text-elicited data. The findings suggest that variance in children's reading comprehension can be predicted by grammatical and lexical features extracted from productive written and spoken language. The study highlights how ML methodologies can enable nuanced examination of the language processes associated with reading comprehension. Educational Impact and Implications StatementMany studies have explored the association between children's reading comprehension and their vocabulary and grammar skills. Usually, such studies require multiple-choice or other forms of selected response measures. This study applied natural language processing to children's open-ended speech and writing, extracting 260 different vocabulary and grammar features. Using machine learning, we found that these features can predict some of the variation in children's reading comprehension, with a slightly stronger relationship found with their writing than with their speech. Among the features that best predicted reading comprehension, grammatical features were more prevalent in children's speech and vocabulary features were better predictors in their writing. Other less well-studied features such as sentiment and word ambiguity also played predictive roles. Although our sample was limited, the results can be considered a proof-of-concept for a novel method of assessing the nuances of children's productive language and literacy skills, as well as understanding how such skills relate to global competencies such as reading comprehension.

Student-Generated Questions in Literacy Education and Assessment

“…All measures were developed as part of a larger research project that assessed students' oral language, literacy, and learning beliefs (Sinclair et al, 2021). The measures used in this study are a questionnaire that asks about students' demographics and learning orientation, one writing task, and two reading comprehension measures with an associated question-generating task each.…”

What explains Macau students’ achievement? An integrative perspective using a machine learning approach (¿Cuál es la explicación del rendimiento de los estudiantes macaenses? Una perspectiva integradora mediante la adopción del enfoque del aprendizaje automático)

“…It can detect significant relationships, trends, patterns, exceptions and anomalies that would otherwise go unnoticed (Sumathi & Sivanandam, 2006). Machine learning approaches have been shown to be applicable to the educational context and could help educators make evidence-based interventions accordingly (Chen et al, 2019;Kiray et al, 2015;Sinclair et al, 2021;Wang et al, 2022aWang et al, , 2022b.…”

“…El aprendizaje automático puede detectar relaciones significativas, tendencias, patrones, excepciones y anomalías que, de otro modo, podrían pasar desapercibidas (Sumathi & Sivanandam, 2006). Se ha demostrado que la aplicación del aprendizaje automático puede resultar útil en la investigación educativa y podría ayudar a los educadores a diseñar intervenciones basadas en la evidencia (Chen et al, 2019;Kiray et al, 2015;Sinclair et al, 2021;Wang et al, 2022aWang et al, , 2022b.…”

Machine learning for spelling acquisition - How accurate is the prediction of specific spelling errors in German primary school students?

“…Even though ML as an analysis tool in research has been embraced by many fields of behavioural sciences -such as psychology (Stachl et al, 2020) or the health sciences (Chen, Liu, & Peng, 2019) -the use of ML models in the educational sciences has to date been much sparser. This is currently changing as we observe an immense increase in available digital data on all levels of the educational system (Jarke & Breiter, 2019), which has made the use of dataintensive models much more feasible during the last decade; innovative research on reading and writing acquisition has resulted in a variety of ML-based models and tools, such as for the prediction of reading comprehension through lexical and syntactic features (Sinclair, 2020) or readability formulas (François & Miltsakaki, 2012).…”

“…As a result, a significant proportion of students have insufficient spelling skills at the end of primary school (Stanat et al, 2017(Stanat et al, , 2022. Some ML approaches for reading and writing have already proven to be useful in supporting teachers to individualise students' learning (e.g., Sinclair, 2020). However, these approaches are mostly limited to demographics that are already proficiently literate.…”