Friday, November 20, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News Technology and Engineering

Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices

November 20, 2026
in Technology and Engineering
Denise Maddox
By Denise Maddox Scienmag Editorial Profile - Mechanical Engineering
Reading Time: 6 mins read
0
Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices

Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

Educational researchers and higher education analysts face persistent challenges when attempting to evaluate student learning across broad knowledge domains rather than isolated course grades. While learning analytics investigations frequently address questions concerning the separability of knowledge areas or how attainment in one domain forecasts performance in another, registrar offices do not maintain student records in competency-aligned structures. Instead, academic institutions export transcript records as flat grade files where each row documents a single course attempt. These flat exports frequently display conflicting column titles, bilingual or mixed-language course names, repeated attempts, administrative non-numeric grades, and elective offerings outside designated competency boundaries. Historically, educational analysts converted these exports manually using desktop spreadsheets, introducing unrecorded choices regarding course categorization, retake filtering, and student threshold criteria that severely undermine analytical reproducibility.

To establish a transparent, deterministic standard for this data preparation step, researchers Tarık Küçükdeniz and Tuncer Can introduced CurricuMap, an open-source Python package and command-line utility described in the journal SoftwareX. Designed as a library-first pipeline, CurricuMap formalizes course-to-domain mapping by translating human decisions into declarative, versioned YAML configuration files governed by strict schema validation. Rather than operating as an automated black-box algorithm or introducing novel predictive classification models, the software treats the assignment of courses to academic competencies as an explicit substantive judgment that domain specialists must record, cite, and evaluate. The pipeline deterministically executes registrar cleaning procedures, provides per-course provenance records tracing how every matrix cell was derived, and computes rigorous audit statistics assessing the coverage and psychometric reliability of the assembled student-by-domain matrices.

The underlying software architecture comprises six dedicated Python modules designated as taxonomy, classify, prepare, audit, synth, and cli. Operating through an importable function design wrapped by a thin command-line interface, the package ingests raw comma-separated value or Excel files into a standardized tabular format containing required fields for student identification numbers, course titles, and numeric marks. An optional external run-configuration file permits researchers to map idiosyncratic institutional column labels onto the target schema, define explicit policies for resolving repeated enrolments, and specify how missing course grades are handled during mathematical auditing. Because these ingestion and cleaning policies reside in standalone configuration files, investigators never edit or alter source registrar files directly, safeguarding the raw historical records against accidental manual corruption or undocumented spreadsheet modifications.

During the classification phase, CurricuMap evaluates course titles against the ordered rules and substring keywords defined in the taxonomy file, prioritizing exact title overrides before evaluating sequential pattern expressions. Crucially, the software records granular derivation records for every single title, documenting whether an assignment stemmed from an override, a regex pattern, an optional fuzzy string match, or remained unmatched. Courses failing all rules are flagged explicitly in audit reports rather than assigned default domains, while multi-rule matches trigger ambiguity warnings for human inspection. The subsequent preparation module resolves retaken courses by programmatically retaining the maximum mark, computing the arithmetic mean across attempts, or selecting the latest grade by date. It then drops non-numeric sentinel marks, filters out students who fail to meet a required minimum domain coverage, and aggregates individual course scores into wide-format and long-format student-by-domain matrices using means or medians.

To verify the analytical integrity of the resulting matrices, the audit module evaluates domain coverage alongside descriptive metrics and psychometric consistency. CurricuMap calculates Cronbach’s alpha reliability coefficients using the constituent courses within each domain as analytical items evaluated across the cleaned grades of the cohort. Recognizing that students rarely complete identical course combinations, the audit tool handles missing course enrolments by substituting course mean scores, which systematically lowers alpha as the proportion of missing observations expands, or through optional listwise or pairwise procedures. To avoid false claims of measurement consistency, the reporting engine highlights underlying assumptions of tau-equivalence, cautions that alpha indicates an internal consistency lower bound rather than confirmed unidimensionality, and deliberately strips color status flags whenever reliability rests on fewer than 30 students, where even unrelated grades can reach apparent reliability coefficients of .70.

Because student academic transcripts represent sensitive personal data under statutory frameworks such as the European Union General Data Protection Regulation and Türkiye’s Personal Data Protection Law No. 6698, CurricuMap implements strict privacy-preserving operational safeguards. The software functions entirely in local compute environments without transmitting records over external network connections. Furthermore, the package incorporates a seeded synthetic transcript generator that synthesizes realistic mock cohorts without reading or processing confidential institutional records. The generator injects functional irregularities encountered in live university environments, such as administrative sentinel zeros, dropped enrollments, and course retakes, strictly parameterizing cohorts using random seeds, target cohort sizes, and the active taxonomy specification. Because these generated records derive from functional parameters rather than confidential microdata, the resulting datasets remain fully anonymous under international privacy laws while allowing peer reviewers and researchers to run continuous integration workflows and reproduce pipeline behavior precisely.

The computational performance and memory consumption of the pipeline were benchmarked on an Apple M2 Max processor running macOS with 32 gigabytes of random-access memory across synthetic cohorts ranging from 60 to 100,000 students and course catalogues of 15 and 150 unique titles. For small datasets of 60 students and 15 courses generating 872 grade rows, the pipeline completed execution in a median time of 0.51 seconds with a peak memory footprint of 81 megabytes, with run duration dominated primarily by Python interpreter startup overhead. Scaling testing demonstrated linear efficiency: a mock institution comprising 1,000 students and 150 courses totaling 145,912 records executed in 0.71 seconds utilizing 124 megabytes of memory. At the upper testing boundary, an expansive historical archive containing 100,000 students and 1,469,156 individual grade records was fully ingested, cleaned, mapped, and audited in 3.19 seconds using 412 megabytes of peak memory, demonstrating that institutional-scale multi-decade transcript analyses can be executed locally on modest personal computers.

To evaluate mapping precision against real-world institutional complexities, the authors validated the bundled national English Language Teaching taxonomy against 143 distinct course titles extracted from the 2017 to 2026 registrar database of İstanbul University-Cerrahpaşa. This historical export contained titles formatted in uppercase with varied abbreviations and numerals, comprising 128 titles matching national curricula from 2006 or 2018, 10 institutional variant spellings, and 5 unlisted local electives or courses for international cohorts. The taxonomy specification had been authored from curriculum policy documents by an automated language model coding workflow in 85 minutes and frozen prior to validation testing without manual human edits to the rule set. Applied to the institutional catalogue, the frozen nine-rule taxonomy correctly assigned all 143 titles to their established reference domains, resolving 10 compound titles through rule ordering while recording required ambiguity flags in the provenance output.

Comparative evaluations against alternative classification techniques demonstrated that rule-based specifications achieve competitive performance while preserving strict auditability. When supplied with labelled official titles, character n-gram logistic regression models, E5 text embeddings, string similarity matching, and large language model classifiers reached overall accuracy rates between 98.6% and 100.0%, exhibiting no statistically significant performance difference compared to CurricuMap’s rule execution (p ≥ 0.500 under exact McNemar tests). Conversely, exact-title lookup matching achieved only 89.5% accuracy (p < 0.001) due to its inability to generalize to registrar spelling variants. When evaluated zero-shot without labelled titles, embedding similarity achieved 68.5% accuracy, an open-weight 12-billion parameter model reached 91.6% accuracy, and a proprietary frontier model achieved 98.6%, with several model runs yielding inconsistent classifications across identical queries or struggling with historical course categorizations that required institutional documentation to resolve.

The investigators emphasized distinct methodological limitations inherent to the study design and software implementation. The institutional evaluation was conducted within a single university department adhering closely to national higher education frameworks, meaning that 128 of the 143 evaluated titles represented standard national entries that most supervised systems easily resolve. Furthermore, because rules utilize substring matching, unrelated off-curriculum courses containing standard lexical roots could be assigned to a competency domain rather than flagged as unknown; for instance, a non-program course like Russian Grammar would match the stem for grammar and map to Language Skills. Consequently, researchers must examine the generated provenance tables rather than presuming that the absence of unmapped flags guarantees comprehensive assignment validity, and newly constructed taxonomy specifications should undergo expert curriculum review prior to reuse.

By transforming transcript preprocessing into a versioned, shareable artifact, CurricuMap bridges an operational divide between raw administrative institutional archives and reproducible educational data science. Academic departments operating under shared degree frameworks can exchange validated taxonomy files, incorporating targeted local overrides to account for regional electives while maintaining cross-institutional comparability. Additionally, the declarative configuration architecture allows researchers to conduct multiverse sensitivity analyses, testing whether substantive learning analytics conclusions remain robust when cohorts are subjected to alternative course-mapping conventions or competing retake resolution policies. The core software, accompanying synthetic generators, and comprehensive validation scripts have been deposited in public archives under the MIT license, establishing an empirical foundation for auditable, transparent institutional research.

Subject of Research: Open-source computational pipeline for converting registrar transcript data into auditable competency matrices

Article Title: CurricuMap: an open-source Python pipeline for converting registrar transcript exports into auditable student-by-competency-domain matrices

Article References: Küçükdeniz, T., & Can, T. (2026). CurricuMap: an open-source Python pipeline for converting registrar transcript exports into auditable student-by-competency-domain matrices. SoftwareX, 36, Article 103235. https://doi.org/10.1016/j.softx.2026.103235

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103235

Keywords: learning analytics, CurricuMap, educational data mining, Python software, transcript processing, curriculum mapping, competency domains, data provenance, higher education, psychometrics, Cronbach alpha, reproducible research, synthetic data generation, YAML specification, schema validation, Turkish higher education

Cite Scienmag News

Denise Maddox. (November 20, 2026). Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices. Scienmag. https://scienmag.com/open-source-python-tool-curricumap-converts-university-transcripts-into-auditable-competency-matrices/

Denise Maddox. "Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices." Scienmag, 20 November 2026, https://scienmag.com/open-source-python-tool-curricumap-converts-university-transcripts-into-auditable-competency-matrices/. Accessed 20 November 2026.

Denise Maddox. "Open Source Python Tool CurricuMap Converts University Transcripts into Auditable Competency Matrices." Scienmag. November 20, 2026. https://scienmag.com/open-source-python-tool-curricumap-converts-university-transcripts-into-auditable-competency-matrices/

Tags: competency domainsCronbach alphacurriculum mappingCurricuMapdata provenanceeducational data mininghigher educationlearning analyticspsychometricsPython softwarereproducible researchschema validationsynthetic data generationtranscript processingTurkish higher educationYAML specification
Share26Tweet16
Previous Post

Faces Pain Scale-Revised Shows Superior Consistency After Glaucoma Goniosynechialysis Surgery

Next Post

Circadian Syndrome Linked to Higher Risk of Stroke in Arthritis Patients

Related Posts

Vibration Analysis Tracks Stiffness and Corrosion in Reinforced Concrete Beams
Technology and Engineering

Vibration Analysis Tracks Stiffness and Corrosion in Reinforced Concrete Beams

November 20, 2026
XGBoost Model Accurately Classifies Nanomaterial Composites Using Multivariate Characterization Data
Technology and Engineering

XGBoost Model Accurately Classifies Nanomaterial Composites Using Multivariate Characterization Data

November 20, 2026
Consortium Blockchain Framework Tracks Container State Changes for Cloud Auditing
Technology and Engineering

Consortium Blockchain Framework Tracks Container State Changes for Cloud Auditing

November 20, 2026
Three-Dimensional Compositional Waves Enable High-Strength Hardenable Titanium Alloys
Technology and Engineering

Three-Dimensional Compositional Waves Enable High-Strength Hardenable Titanium Alloys

November 20, 2026
Lubrication Dampens Bearing Vibration While Skidding and Defect Impacts Intensify Motion
Technology and Engineering

Lubrication Dampens Bearing Vibration While Skidding and Defect Impacts Intensify Motion

November 20, 2026
Solar-Activated Bismuth Oxyiodide Nanostructures Rapidly Degrade Pollutants and Track Chemical Breakdown
Technology and Engineering

Solar-Activated Bismuth Oxyiodide Nanostructures Rapidly Degrade Pollutants and Track Chemical Breakdown

November 20, 2026
Next Post
Circadian Syndrome Linked to Higher Risk of Stroke in Arthritis Patients

Circadian Syndrome Linked to Higher Risk of Stroke in Arthritis Patients

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Career Adaptability Linked to Higher Psychological Well-Being Among Iranian Healthcare Workers
  • University of Konstanz Launches Toxic Response Research Group With DFG Funding
  • Salivary miRNA Panels Show Promise for Oral Cancer Detection
  • Geological Injection Limits Could Reshape China’s Heavy Industry Climate Transition Pathways

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading