Revealed Learning Interest as a Curriculum-Alignment Signal: data, code and validation records for a three-layer study of Jordanian ICT education
Year: 2026
Type: dataset
Abstract: Data, code and validation records accompanying the manuscript "What Arabic-Speaking Learners Actually Seek, and How They Feel About It: A Three-Layer Comparison of Curricular Provision, Revealed Learning Interest and Learner Affect in Jordanian ICT Education" (supporting a manuscript on curriculum alignment in Jordanian ICT education). The study builds three independently collected layers over one controlled vocabulary, the ESCO v1.2.1 digital skills collection (1,284 concepts), and joins them on conceptUri so that an Arabic course description, an English video title and an Arabic comment naming the same skill resolve to the same concept without any translation step. Layer 1 is 988 course descriptions from twelve Jordanian ICT programmes; layer 2 is 106,737 YouTube videos from 1,005 channels retrieved through 48 Arabic learning-intent queries and channel harvesting; layer 3 is 434,951 public comments, from which 30,491 aspect-bearing Arabic segments are scored by a dialect-aware transformer. The deposit contains: the Layer 1 curriculum corpus; the Aho-Corasick skill matcher and its aspect lexicon and blocklist; the byte-exact 48 seed queries; the concept-level and skill-level derived tables behind the reported results; the hand-coded validation records for the matcher, for aspect assignment and polarity, for the second coder, and for the difficulty disambiguation rule; the cell-by-cell scripts that compute every validation statistic reported in the manuscript and its Online Resource, including the corpus-weighted confusion matrices, the channel-cluster bootstrap intervals, the attention sensitivity variants and the restricted-reading comparison; the four collection and analysis notebooks; and the matplotlib sources for all six figures with the Springer EPS and 600-dpi TIFF export script. Third-party text is not redistributed. The comment corpus and the raw video metadata are not included, the comment-level validation records carry labels without the underlying comment text, and advertisement excerpts have been removed from the abandoned vacancy pilot. No commenter identity was ever stored. The collection code is included so that a reader holding a YouTube Data API v3 key can regenerate both corpora. See README.md inside the archive for the file manifest, a table mapping each reported number to the file that produces it, and docs/DATA_DICTIONARY.md for column definitions.
Authors: Hmoud, Aladeen, Altalib, Raya Ahmad Hasan
Institutions: Yarmouk University, German Jordanian University
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7204705853