Since ECER 2024 in Nicosia, the interactive webapp EduTopics: ECER provides interested users access to information on the contributions to ECER since 1998 (Christ et al., 2025a). In 2025, the webapp EduTopics was updated with additional functions based on feedback from the audience in Nicosia, general users and the national representatives and link convenors of the EERA networks. In 2025, EduTopics provided the basis for several contributions at ECER 2025 in Belgrade, was presented at EARLI 2025 in Graz and was used for various publications as well, we were able to gather additional feedback on its design and functionality. Meanwhile, the presentations beyond ECER with other (inter-)national organizations opened up new collaborations regarding the inclusion of data of other conferences on educational research in EduTopics. The resulting transformation from EduTopics: ECER to a cross-national, cross-conference webapp simply called EduTopics, its new design and functions are the focus of this presentation.  

EduTopics contains data and manipulable visualisations on the authors, countries of affiliation, EERA-networks and themes for n > 30,000 contributions from ECER (exact number of contributions from other sources are to be determined). The themes of the contributions were identified by applying the computer linguistic machine learning approach topic modelling – more specifically latent Dirichlet allocation (LDA, Blei et al., 2003) – which identifies word and document clusters in large literature corpora (Blei & Lafferty, 2009; Vayansky & Kumar, 2020). A hierarchical approach to topic modelling enables the identification both of broader, overarching supertopics and more detailed smaller subtopics similar to super- and subcategories in qualitative content analysis. By aggregating the resulting (sub- and super-)document-topic-weights in regard to the aforementioned covariates (i.e. authors, countries, networks), the thematic foci and trends of authors, countries and networks can be identified, visualized, interpreted and compared. Additionally, users can upload own singular texts or entire corpora to determine the latent topics of the text(s) based on the topics of ECER, the EERA-network most relevant to the uploaded text and contributions with similar content.  

The general development process of the webapp is ongoing but each version is uploaded separately to ensure the replicability and reproducibility of the results of research papers utilizing the app (for 2024 see Christ et al., 2024; for 2025 see Christ et al., 2025b). New functions of the 2026 version of EduTopics (see Christ et al., 2026) include:

•    A new global settings menu to apply consistent filters to the entire conferences and their dataset. Options for filtering include for example years and data source (i.e. conference). 
•    The option to search the super- and subtopics for specific terms and visualize the term-topic-weights β.
•    A search function to rank the contributions regarding multiple super- and subtopics. This enables users to determine documents which are relevant to a mixture of specific topics such as contributions with high document-topic-weights γ for a specific content-related AND method-related topic. 
•    The option to show the specific topic related measures of single contributions including the bibliographic covariates. 

Additionally, future developments are discussed such as the development of a community platform to enable a community-led approach to the qualitative aspect of topic labelling or the development of a data and modelling pipeline to include additional data sources.

The central method of EduTopics is LDA topic modelling. The themes in the corpus were identified via a hierarchical, nested topic modelling approach by simultaneously modelling (1) a model with a large number of topics (between k > 300) capturing smaller thematic clusters and (2) a broader model with a smaller number of topics (around k = 50) representing broader central themes within the corpus. The similarities between and within models are determined by aggregating various distance and similarity measures both on the document and word level (Aletras & Stevenson, 2014; Blair & Mulvenna, 2020). This similarity measure can then be utilized to determine the nesting of the subtopics (of model 1) within the supertopics (model 2) as well as the pairwise relationship of topics within one of the two models. The results of both topic models and the results of the frequency analyses of the covariates (i.e. author, country, EERA network, year) were aggregated via their categorical values/mean-topic-weights which in turn resulted in for example author-topic-weights, EERA-network-country distributions and various trend analyses. All results are available as interactive and manipulable figures and tables, allowing for example filtering of the data or downloading relevant documents in tabular formats. Exemplary results to be presented include differences in thematic foci and trends between the different data sources or the integrated mapping functions both for single texts and literature collections.

This contribution provides in-depth information into the current and future development of the app EduTopics. In addition, it presents use-cases of the app specifically and for natural language processing and machine learning in general for educational research purposes. It also highlights how the app and the utilized methods in general open up new ways for researchers to access data, which may not have been possible before with traditional methods such as qualitative content analysis. The discussion will focus on potential future challenges for modelling multilingual literature corpora and for deploying the aforementioned community-platform and the data pipeline. The paper aims to elicit constructive feedback from participants to support the ongoing development and improvement of EduTopics.

Aletras, N., & Stevenson, M. (2014, April). Measuring the similarity between automatically generated topics. In Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, volume 2: Short Papers (pp. 22-27). Blair, S. J., Bi, Y., & Mulvenna, M. D. (2020). Aggregated topic models for increasing social media topic coherence. Applied intelligence, 50(1), 138-156. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. Journal of machine Learning research, 3(Jan), 993-1022. Blei, D. M., & Lafferty, J. D. (2009). Topic models. In Text mining (pp. 101-124). Chapman and Hall/CRC. Christ, A., Röschlein, J., & Schindler, C. (2026). EduTopics: ECER. https://dipf-lis.shinyapps.io/EduTopics [placeholder url] Christ, A., Röschlein, J., & Schindler, C. (2025a). EduTopics: ECER: A webapp for interactive visualisation and exploration of ECER contributions since 1998. European Educational Research Journal, 14749041251405570. Christ, A., Röschlein, J., & Schindler, C. (2025b). EduTopics: ECER 2025. https://dipf-lis.shinyapps.io/EduTopicsECER Christ, A., Röschlein, J., & Schindler, C. (2024). EduTopics: ECER - An interactive app for visualising and exploring the contributions of ECER-conferences for 1998 to 2024. https://dipf-lis.shinyapps.io/EduTopicsECER2024 Vayansky, I., & Kumar, S. A. (2020). A review of topic modeling methods. Information Systems, 94, 101582.