ASEANLex: A Traceable Retrieval and Competency-Matching Prototype for ASEAN-Oriented Legal Education
DOI:
https://doi.org/10.63593/SLJ.2026.09.07Keywords:
legal education, information retrieval, BM25, evidence abstention, competency matching, source provenance, explainable recommendationAbstract
Legal-education platforms that combine document retrieval and competency-based career guidance must distinguish topically similar teaching material from reliable legal authority and differentiate high suitability scores from mandatory eligibility. This paper presents a source-code-grounded engineering diagnostic of ASEANLex, a TypeScript/Express prototype comprising a local lexical index, a coverage-based evidence gate, a five-dimension competency rubric and an explanatory demonstration-job matcher. The development collection contains 196 learning records. Five retrieval configurations were compared on 40 corpus-informed reference prompts and 22 researcher-authored unsupported probes. Ungated BM25 placed a provisional reference within the first three results for 32 of 40 prompts, against 25 of 40 for token overlap. A gate at 0.34 reduced unsupported-probe acceptance from 18 to seven of 22 while retaining support for 39 of 40 reference prompts; a 0.60 gate reduced unsupported acceptance to four but retained only 21 reference prompts. Separately, 112 constructed learner–job scores and a 45-case qualification overlay exposed five soft-pass cases without imposed mandatory-language evidence. These are reproducible development diagnostics, not expert-judged accuracy, improved student learning, or observed recruitment outcomes. We specify a blinded annotation and provenance-validation pathway for subsequent evaluation.
