Teaching a computer to read 1.5 million Peruvian parish records — and the two Texas State master’s students behind it

In the Peruvian Andes, Catholic parishes recorded baptisms, marriages and burials by hand from the 18th century into the early 20th. More than 1.5 million entries, scattered across thousands of pages, with no practical way to connect them. A team at Texas State is teaching a computer to do it.

But the reason we are running this story is not Peru. It is the last number in that row.

The hard part is not reading the handwriting

The project is called "Sondondo Valley. Population and Territory in Rural Andean History." The Texas State side is led by Dr. José Carlos de la Puente, professor in the Department of History, who grew up in Lima.

The real difficulty is not character recognition. It is recognising one person when they surface three times, in three kinds of register, decades apart, with their name spelled three different ways. A computer treats each appearance as somebody new.

And each entry holds more than one person:

On top of that: a different hand for every parish priest, no settled spelling, and centuries of transcription error.

Two master's students, and neither studies history

This is the part worth reading if you are weighing a master's in the United States. The technical work on the project is being done by two of the university's own graduate students.

One computer scientist, one industrial engineer, both working on colonial Peruvian history. That is a concrete answer to the two questions prospective graduate students ask most: do you actually get to do research, or only coursework — and is this degree of any use outside its own narrow field.

And the sharpest methodological point comes from the industrial engineer:

The skill they are learning has a name, and a market

Here is what the original article does not say, because it was written for alumni readers in the United States. For the readers of this site it is the most useful part.

The technique is probabilistic record linkage. Instead of a hard rule — the names must match exactly — the model computes the probability that two records describe the same person, weighing several fields at once.

Their tool is Splink, an open-source Python library. Worth noting where it comes from: Splink was built by the data-linking team at the UK Ministry of Justice to join administrative datasets across the justice system. This is not an academic toy.

Is $23,400 a lot or a little

Texas State funded this phase with a $23,400 grant from the Research Enhancement Program, run by the Division of Research. That figure on its own tells you nothing, so we looked up the programme's rules.

So: near the top of the programme. But this is an internal university fund, not a federal one, and internal funds are seed money — enough to carry a line of work far enough to produce results worth taking to a larger external funder.

Keep that yardstick and the next "awarded $X" headline you read, from any university, prompts the right question: internal or external money, and what is the ceiling.

The digitisation itself began in 2019, funded by the French Institute of Andean Studies. In 2023 the project team directed student assistants at the University of California–Santa Barbara in processing some of the records into a database. What Texas State is doing is the next stage.

What exists, and what is still a plan

Worth separating, because it is easy to skim this and assume there is already a site to search.

How you would get a place like that

Plainly first: admission to a master's programme does not come with a research post attached. This is two people, on one specific grant, with one specific professor. But the way to get close is real and doable:

• Read the department's research pages before you apply, not after you arrive. Names, live projects and recent publications are all public on the department site.
• Write to faculty about something specific you can do, not about your enthusiasm. "I have cleaned datasets in pandas and used fuzzy matching" lands harder than "I am passionate about research".
• Do not rule out a project because it sits outside your field. Neither student here studies history. What they brought was method, not subject knowledge.
• Ask the international student office about on-campus work limits before accepting any position — the rules differ by status.

See also: graduate admissions, the programme list, research at Texas State.

Sources: TXST digital history project unlocks centuries of Andean history, Matt Joyce, Hillviews, 18 September 2026 (all quotations, names, dates and the grant figure) · REP Guidelines, Division of Research, Texas State (award ceilings) · Splink, UK Ministry of Justice (the record-linkage tool) · Department of History, Texas State. The connection drawn between record linkage and work in financial services is our own, not the university's. The original article also carries photographs of the parish registers and of the Andes; those are copyright the French Institute of Andean Studies and Evelyne Mesclier and are not reproduced here — see them at the source.