Skip to content
Multigrid

CVs, certificates and identity documents

Papers that assert who someone is or what they have done, including the machine-readable zones designed to be parsed.

The documents in this cluster all make a claim about a person, and they divide sharply on whether anybody standardised the claim. A passport carries a machine-readable zone specified to the character position, with check digits you can recompute; a CV carries whatever its author felt like typing. Treating those two as the same extraction problem is the mistake that produces a pipeline nobody trusts. On one side of the line an extraction can be proved wrong by arithmetic, and on the other the only thing you can do is make the ambiguity explicit in the schema.

So these pages are organised around what the document gives you to check against. Where a standard exists — ICAO Doc 9303 for travel documents, AAMVA for the barcode on a North American driver’s licence, an ISBN-13 check digit in a reference list — the page shows the arithmetic and treats validation as part of extraction rather than as a separate review step. Where none exists, it shows the specific shape of the ambiguity: the overlapping contract role that invents an employment gap, the degree name that means different things in two countries, the field that is blank rather than filled in. The generic machinery of schema design, confidence and review lives elsewhere in the library; this cluster is the document-specific layer on top of it.

Extracting Author Affiliations From a Scientific Paper

The author-to-affiliation join is carried by superscript markers, which are the first thing OCR destroys, and the fix is usually not to parse the PDF at all.

8 min read

Extracting Methods and Sample Size From a Scientific Paper

A paper states several different values of n and all of them are correct, so the schema question is which n a field is supposed to hold.

8 min read

Extracting a Bibliography Into Structured Citation Records

A worked two-stage parser that segments a reference list into individual entries before parsing any of them, because the segmentation is the part that fails.

9 min read

Extracting DOIs and ISBNs From a Reference List

Both identifiers arrive in several surface forms, and one of them carries a check digit you can recompute to prove the extraction is right.

8 min read

Extracting Work History From a CV Into Structured Fields

A role-level schema that survives an open end date, a promotion written as one entry, and two jobs held at the same time.

9 min read

Extracting Skills and Certifications From a Resume

Skills and certifications look like one section and are two different data types: one has no identity and the other has an issuer, a number and an expiry.

8 min read

Extracting Education History From a CV, Degree by Degree

The degree name and the field of study are different fields that most CVs run together, and normalising across countries needs a published level classification rather than a guess.

8 min read

Extracting Employment Gaps From a Resume's Date Ranges

A worked interval-merge over extracted role dates, with the month-rounding rule stated and the naive scan that invents gaps shown failing.

9 min read

Extracting Fields From a Job Application Form

A form mixes free text with checkbox state, and the difference between blank, N/A and absent is three answers that a two-valued schema collapses into one.

8 min read

Extracting a Passport's Machine-Readable Zone Into Structured Fields

The MRZ is specified to the character position by ICAO Doc 9303 and carries four check digits, so a correct extraction can be proved rather than trusted.

10 min read

Extracting Data From a Driver's License Across Different State Formats

The printed face varies by jurisdiction so template matching fails, but the PDF417 barcode on the back is standardised by AAMVA and carries the same element identifiers everywhere.

9 min read

Extracting Fields From a National ID Card

Bilingual cards print the same field twice in two scripts, and the schema has to decide which one is authoritative before it can store either.

9 min read

Extracting Data From a Birth Certificate

Short-form and long-form certificates carry different field sets, and the embossed registrar seal sits on top of the text a schema most wants to read.

9 min read

Extracting Fields From a Marriage Certificate

Two parties occupy structurally identical field blocks, so any extraction that depends on position will eventually swap them and produce a well-formed wrong record.

8 min read

Extracting Data From a Diploma or Degree Certificate

Diplomas are typeset to be ceremonial rather than legible, with Latin honours, Roman numeral dates and institution-specific degree names that need a per-institution glossary.

8 min read

Extracting Fields From a Professional License Certificate

An expiry decision is only meaningful with the date it was made recorded beside it, and the certificate is never the authority on whether a licence is currently valid.

9 min read

Other topics