Introduction

Worldwide, technology has rapidly innovated over the last several decades. With the advance of technological tools and systems in society, information has become easily accessible. Within this time of new and rapid transitions, it is necessary to understand the trends of using technology to teach mathematics to students with disabilities, who are underrepresented in science, technology, engineering, and mathematics (STEM) fields.

Purpose. To analyze emerged topic trends in technology use in mathematics nor explored the evolution of research topics between 1980 and 2021, applying text-mining approaches.

To present the topics and trends in technology use for teaching mathematics to students with disabilities. Word network and topic modeling in a text-mining approach were implemented to examine the evolution of topics and associations among keywords in publications over the last 42 years.

Method

Inclusion Criteria. (a) the target participants of the studies were students with disabilities in K to 12 grades; (b) the focus of studies was on teaching mathematics using technology; (c) studies were journal articles or dissertations published in English between 1980 and 2021; (d) studies reported title and publication year with abstract.

Search Strategy. Applied the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021) for searching journal articles and dissertations publisehd in English via an electronic database search of ERIC, Web of Science, Academic Search Complete, Education Source, APA PsycInfo, and MEDLINE, resulting a total of 10,882 studies (initial search results).

Extraction of Bibliographic Data. Exported the search results as a bibliographic citation file (RIS) through the university library’s EBSCOhost Collection Manager. Then, the researcher imported RIS file to the EndNote (EndNote Team, 2013) library, manually detected duplicates, and removed studies that do not meet inclusion criteria. This procedure resulted in 488 studies for text mining analysis. The dataset is also available through textminingR R package (Shin, 2022).

Text Pre-Processing. To make the textual data appropriate for the algorithm, the researcher has pre-processed textual data in three steps: (a) constructing a corpus by selecting a text column in the dataset (abstract) and combining texts with document-level variables (publication year); (b) constructing a token object by segmenting complex text into smaller words; (c) constructing a document-feature matrix, displaying the frequencies of features (i.e., tokenized words) for each document.

Creating Customized Dictionary and Stop word lists. To avoid duplicating the exact words, the researcher sequentially processed two different dictionary objects (224 words list for the first and 166 words list for the second dictionary) with matching patterns of wildcard expressions and calculated inverse document frequency (idf) of each word to create the customized stop words list. This process was completed through quanteda R package (Benoit et al., 2018). The lists of extended acronyms, customized dictionaries, and stop words list objects are available in the online repository (Shin et al., 2022).

Identification of the optimal number of Topics. Based on the goodness-of-fit results of models with topic numbers ranged from 5 to 30, topic number 15 showed relatively low residuals, high semantic coherence, a maximized lower bound, and a high held-out likelihood.

Structural Topic Modeling. Based on the generative process for each document, a topic is predicted based on a clustered list of words where each word has a probability of belonging to a topic (β). Document (d) consists of a mixture of topic K where the sum of topic proportions per document (θ) is one (Roberts et al., 2019). Considering the continuous change in topic prevalence over time, the proportions of topics over time were examined via the estimateEffect() function of the stm R package (Roberts et al., 2019).

To estimate the average topic proportion for each 15 topics, a cubic spline model with three knots in 1990, 2000, and 2010 was applied, using the following formula: log(θ) = b0 + b1(year.c) + b2(year.c)^2 + b3(year.c)^3 + b4(year.c > 10)(year.c – 10)^3 + b5(year.c > 20)(year.c – 20)^3 + b6(year.c > 30)(year.c – 30)^3 + ε, where year.c is the centered at the first publication year of 1980. The centered publication year was processed as 1 (True) if year.c > 10, 20 or 30, and 0 (False) otherwise. The researcher-developed web tool is available at https://mkshin.shinyapps.io/textminingR.

Results

As shown in (Figure 1), a total of 488 studies were published in English between 1980 and 2021 in teaching mathematics using technology for students with disabilities in K to 12 grades; 85% were journal articles (n = 416), and 15% were dissertations (n = 71).

Highest (Top 10) Word Probabilities for Each Topic. As shown in (Figure 2), the highest word probabilities were in a wide range for 15 topics in the special education technology to teach mathematics for students with disabilities. The labeling of each topic was based on the highest word probability within each topic and the distribution of probabilities. In case of the topic, “Computers”, the dominant word probability was on the keyword of “computer.” However, in case of “Anchored instruction,” “Self modeling,” and “Virtual impairment,” wider range of word probabilities and terms represented the topic.

Evolution of Topic Prevalence Between 1980 and 2021. As shown in (Figure 3), there was various and rapid change in many topics over the last 42 years. The topics of “Computer-assisted instruction” and “Mathematics achievement” showed relatively higher average topic prevalence in the earlier decades in 1980s and 1990s. Although the degree of topic prevalence for each topic was different, the research on “Instructional sequence,” “Mobile apps,” “Online learning,” and “Virtual manipulatives” have shown rapid increase since 2010.

Conclusion

Understanding the trends and associated keywords can help practitioners and researchers review the patterns of research interests in the field. Reflecting the frequently used words/topics and their associations related to technology in mathematics instruction for students with disabilities, educators can educators the inspect of the past instructional patterns and predict future changes. Future research can further focus on how the topic proportions differ by other document-level variables, inducing a time variable as before and after the COVID-19 pandemic, and disseminate the recent changes to the public.

Contact. mikyung.shin@wtamu.edu Collaborators. Min Wook Ok, Sam Choo, Gahangir Hossain, and Diane P. Bryant Reference. Shin, M. (2022). textminingR: Text mining workflow tools (R package version 0.0.1) https://github.com/mshin77/textminingR