Transcript Parser: Translating Academic Records into Skills and Opportunities

An AI-powered transcript parser that reads college transcripts, extracts the institutions, courses, credits, grades, and degrees they contain, and maps them to CTDL records in the Credential Registry — returning globally unique identifiers (CTIDs) and the linked data behind them, including the skills and competencies each course teaches.

37 million Americans have completed college courses without earning a credential. Their transcripts are static records of learning that employers and institutions cannot easily interpret. By connecting transcript contents to the Credential Registry's transparent, machine-actionable descriptions, this tool turns academic records into dynamic skills profiles — supporting skills-based hiring, faster transfer credit evaluation, and clearer paths back to credential completion.

A collaboration of Credential Engine and AACRAO.

What it does

Given a folder of transcripts (PDF, including scanned documents), the parser:

1. Extracts the issuing institution, transfer institutions, every course (subject, number, title, credits, grade, term, level, transfer status), degrees awarded, and transcript dates — using Claude with a structured schema. Learner PII is excluded by default.
2. Matches institutions to Registry organizations, handling abbreviations, campus qualifiers, and district/system naming.
3. Matches courses and credentials to the institution's published Registry records (courses, learning opportunities, programs, degrees) by course code and title similarity, with a confidence rating on every match.
4. Builds a skills profile: competencies linked to matched courses (ceterms:teaches) are retrieved, deduplicated, and traced back to the courses that taught them.
5. Reports everything as an HTML report (matches, non-matches, confidence, skills), plus JSON and CSV for downstream processing — including which institutions have no published Registry data, a useful view of Registry coverage.

Quick start

python -m venv .venv
.venv\Scripts\Activate.ps1
pip install -r requirements.txt

$env:ANTHROPIC_API_KEY = "<Anthropic API key>"
$env:CE_REGISTRY_API_KEY = "<Credential Registry Search API key>"

python -m transcript_matcher <folder-of-transcripts> --out output

Open output/report.html for the match report. Useful flags: --limit N (first N files), --extract-only (no Registry key needed), --env sandbox (sandbox Registry), --model (default claude-opus-4-8).

python -m transcript_matcher.probe "<institution name>" tests organization matching directly.

Privacy

Transcripts contain FERPA-protected information. Extraction is instructed to discard learner PII (names, IDs, birth dates), and no transcript documents or run outputs belong in this repository. Note that transcript files are transmitted to the Anthropic API for extraction; confirm current data-handling policy before processing non-redacted documents.

Roadmap

See ROADMAP.md for the phased plan: richer academic data and PII-inclusive normative output, calibrated confidence scoring for registrar use, conversational access via MCP, credential gap analysis, and transcript verification/trust.