The OER Observatory aggregates and displays OER-related news and content from third-party sources worldwide. This content is not produced by UNESCO or the OER Dynamic Coalition. All rights and responsibility for the original content remain with the respective authors and publishers. The inclusion of third-party content in the Observatory does not imply endorsement by UNESCO or the OER Dynamic Coalition. Users can access the original source through the link provided with each item.
Researchers have developed a semantic curriculum mapping system, known as CCMAP, designed to automate the integration of UNESCO's Greening Curriculum Guidance into national educational practice. The system addresses the substantial localization gap that often exists between global policy frameworks for climate change education and the specific, day-to-day requirements of local classroom lesson planning. By leveraging a reproducible computational pipeline, the software inventory translates abstract global learning outcomes into actionable entry points within national curriculum statements, providing teachers with a structured bridge to locally available educational resources. This development marks a transition from qualitative, expert-driven curriculum alignment toward a computational approach that enables transparent and inspectable pedagogical design, aiming to support teachers who must incorporate climate change content into already overburdened existing subjects.
The central design challenge recognized by the research team is that the primary bottleneck for climate change education implementation is not teacher motivation, but the significant workload associated with curriculum design. While global organizations like UNESCO provide clear articulation of key ideas and learning outcomes across cognitive, social, and emotional domains, teachers must base their lesson plans on national standards, specific grade levels, and subject domains. Attempting to bridge this gap manually for every teacher and subject is widely considered infeasible. To address this, CCMAP treats national curriculum statements as the primary entry point for lesson planning rather than global sustainability topics, grounding the search process in the exact material teachers interact with daily. This design ensures that the software assists practitioners in the immediate context of their professional work rather than adding a layer of abstract pedagogical requirements.
At the technical core of the system is a two-stage semantic pipeline. In the first stage, a teacher-selected curriculum statement is matched against a localized database of UNESCO key ideas using precomputed dense text embeddings. These embeddings allow the system to compare the semantic content of a curriculum statement against the framework's learning outcomes, identifying the most relevant climate-related themes regardless of the specific vocabulary used in the national curriculum. Once a key idea is selected, the second stage retrieves a ranked list of locally available educational resources, such as lesson plans, videos, and articles, based on their similarity to the key idea. By routing recommendations through this intermediate pedagogical layer, the system keeps the rationale behind every resource recommendation visible, encouraging teachers to engage with the learning outcomes before moving to concrete classroom materials.
The system is built as a Flask-based web application and relies on four swappable data layers that are designed to lower the barrier for other countries to supply their own information. These include a global framework layer derived from UNESCO guidance, a local curriculum layer containing the specific learning-content statements of a country's national curriculum, a curated resource index, and a layer of derived vectors used for matching. By using plain CSV files for these layers, the system architecture allows administrators in other regions to replace the Taiwan-specific data with their own national curriculum statements without needing to alter the underlying code. This portability is a key operational property, as it facilitates the use of the localization workflow defined by UNESCO in diverse contexts while maintaining consistency in how the system computes similarity.
To maintain operational efficiency and accessibility, the system performs all corpus embedding computations offline. Runtime cost is strictly limited, as the system only triggers a paid embedding API call when a user adds original free-text input to their search queries. This design choice ensures that the platform remains cost-effective for small public-sector teams, as item-based browsing -- which includes the full two-stage flow from curriculum statement to teaching resource -- can function without any API key. The DataLoader component at the system's core enforces data consistency, ensuring that the dimensions of the embedding matrices and the row counts of the text files align before the application begins serving requests, which guards against the use of stale or incompatible data.
During a sixteen-month production deployment in Taiwan starting in March 2025, the researchers recorded extensive interaction data that demonstrates sustained voluntary uptake by educators. Analyzing over 5,000 interaction events from more than 600 distinct users revealed that the platform's usage pattern aligns closely with professional activities, with the vast majority of engagement occurring on weekdays and during typical school hours. The logs capture clear evidence of the system's utility in cross-domain issue integration, showing that searches are not limited to science but extend across social studies, health, and other subjects where teachers are tasked with infusing environmental content. This usage data provides naturalistic evidence that the curriculum-entry-point design effectively scaffolds the integration of sustainability themes in complex educational landscapes.
The research team also implemented a community quality loop to address the common metadata-quality failure modes found in open educational resource discovery. Users can report issues for each resource, such as broken links, misclassifications, or inappropriate content, and this feedback channels corrections directly back into the index. Over the course of the deployment, multiple curation passes were performed to remove duplicates and resources with incompatible terms of use, demonstrating an iterative approach to maintaining a high-quality, actionable index. This system of issue reporting provides a safety mechanism for providers to object to inclusion, allowing for immediate withdrawal of links while maintaining a transparent relationship between the platform and the organizations hosting the materials.
When analyzing the coverage of the sustainable development framework, the research revealed an important disparity in resource availability. While the system identifies alignments for all learning domains, the 'Society' domain -- which includes topics like climate justice and the political economy of transitions -- is currently the thinnest in the index. These key ideas frequently return a smaller set of resources at lower similarity scores compared to the 'Environment' or 'Economy' domains. The researchers emphasize that this gap profile provides an evidence-based roadmap for future resource curation, rather than a failing of the software itself. By highlighting where curricula are thin on specific themes, the system allows curriculum developers and resource curators to prioritize their efforts effectively, focusing on the contested social and political facets of climate change that teachers struggle to address.
The transferability of the pipeline was validated through tests using unseen data from other systems, including England's national curriculum. In these supplementary tests, the system produced clear correspondences and a consistent gap profile, reinforcing the observation that climate justice and the political-economic aspects of climate change represent the most challenging areas for existing curricula to cover. These results indicate that the similarity matrix is readable in both directions, allowing developers to assess how well their national curriculum captures the breadth of the UNESCO framework. Each of these transferability runs took only seconds and required negligible computational expense, further proving the suitability of the approach for varied international contexts.
Despite these capabilities, the researchers stress that the system provides semantic similarity, which is a proxy for, rather than a guarantee of, pedagogical fit. The software is designed to supply competence-oriented context that helps teachers navigate policy documents, but it is explicitly not a replacement for expert judgment. The authors contend that tools like CCMAP should be evaluated on their ability to facilitate professional decision-making rather than on their capacity to dictate teaching outcomes. As such, future work includes planning for structured user studies that incorporate pre- and post-use questionnaires to better understand the impact of the tool on the planning process, administered alongside an alignment task to gather direct feedback on how practitioners evaluate the machine-produced suggestions.
Privacy and security considerations are integrated into every aspect of the platform's design, adhering to the principles of data minimization. The system does not store names, email addresses, or account data, and it utilizes browser-generated pseudonymous identifiers for all logs. Because even pseudonymous event traces, including verbatim free-text queries, could theoretically be re-identifiable, the researchers maintain a policy of not publishing raw logs. All data availability reports and analyses are conducted using aggregate operational telemetry, ensuring that the platform remains a safe environment for teachers to explore curriculum alignments without the risk of individual tracking or personal data exposure.
In conclusion, the project demonstrates that the complex challenge of localizing a global climate education framework can be successfully packaged as reproducible, open-source software. By combining structured framework translation, curriculum-anchored semantic alignment, and a curated, inspectable resource index, CCMAP provides a concrete worked example for other national contexts. The deployment experience in Taiwan highlights that tools designed to reduce teacher workload while keeping pedagogical control in the hands of educators can achieve continued professional uptake. As countries worldwide face the 2030 target for integrating climate change into their curricula, the researchers suggest that computational alignment tools will become increasingly important for maintaining the coherence and quality of sustainability education in the face of crowded national standards.
Subject of Research: Computational educational technology for curriculum alignment with climate change guidance.
Article Title: CCMAP: A semantic curriculum mapping system for localizing UNESCO's Greening Curriculum Guidance into national curriculum practice
Article References: Chang, Y.-J., Chen, T.-Y., Chang, C.-H., Su, H.-J., Yang, L.-C., Chen, W., & Lin, H.-C. (2026). CCMAP: A semantic curriculum mapping system for localizing UNESCO's Greening Curriculum Guidance into national curriculum practice. SoftwareX, 36, Article 103220. https://doi.org/10.1016/j.softx.2026.103220
Image Credits: AI Generated
Keywords: Climate Change Education, Curriculum Mapping, Educational Technology, UNESCO, Semantic Alignment, Open-Source Software, Pedagogical Design, Teacher Workload, Flask, Sustainable Development Goals, Embedding Models, Curriculum Integration, Natural Language Processing, Climate Literacy, Computational Linguistics, Teacher Education, Educational Resources, Taiwan, Machine Learning, Localized Curriculum