Glossary
What is Resume Parsing?
Resume parsing is the automated extraction of structured data — name, contact info, skills, work history, education — from an unstructured CV file (PDF, DOCX, or image). It turns a human-formatted document into machine-readable fields that a hiring system can filter, rank, and display.
How a modern resume parser works
- 1.File ingestion: PDF, DOCX, or image uploaded.
- 2.Text extraction: OCR for scans/images, native text extraction for digital PDFs/DOCX.
- 3.Section detection: Identify "Experience," "Education," "Skills" blocks.
- 4.Entity extraction: Pull company names, job titles, date ranges, degrees, skill keywords.
- 5.Normalization: Map skills to canonical forms (e.g., "JS" → "JavaScript"), dates to ISO format.
- 6.Structured output: Return JSON with fields the hiring system can use.
Where traditional parsers fail
- ✓Multi-column layouts confuse section boundaries — experience gets mixed with skills
- ✓Non-English CVs mis-identify sections when headers are translated
- ✓Creative formats (infographic resumes) get mangled or rejected
- ✓Skills hidden in project descriptions are missed if not in a "Skills" section
LLM-powered parsing
Modern parsers (including The Scorify) use large language models to read CVs in context, not by layout rules. That means a skill mentioned in a project description counts, a two-column layout doesn't break extraction, and cross-language resumes parse cleanly. Accuracy typically jumps from ~70% (rule-based) to 90%+ (LLM-based) on messy real-world CVs.
What to evaluate in a parser
- ✓Multi-language support (English + at least FR/ES/DE)
- ✓Creative-format tolerance (PDF with images, two-column layouts)
- ✓Skill normalization (handles synonyms, not just exact strings)
- ✓Speed at bulk scale (100+ CVs batched)
- ✓Transparent output (you can see the raw extracted JSON)