Nuestra MemorIA Project will rescue information from human rights archives using artificial intelligence
**June 2024.-** Harnessing the transformative power of artificial intelligence (AI) to manage, analyze, and interpret historical archives from the era of the Chilean dictatorship (1973-1990): this is the goal of the [Nuestra MemorIA Project](https://nuestramemoria.ing.uc.cl/) from the **Millennium Institute for Foundational Research on Data (IMFD)**, led by [Jocelyn Dunstan Escudero](https://www.ing.uc.cl/academicos-e-investigadores/jocelyn-mariel-dunstan-escudero/), faculty member at [DCC](https://dcc.ing.uc.cl/) and [IMC UC](https://imc.uc.cl/); [Juan Reutter](https://www.ing.uc.cl/academicos-e-investigadores/juan-lorenzo-reutter-de-la-maza/), faculty member at [DCC](https://dcc.ing.uc.cl/) and [IMC UC](https://imc.uc.cl/) and director of IMFD; **Camila Diaz**, executive director of IMFD, [Antonia Fonck](https://uc-cl.academia.edu/antoniafonck), [Instituto de Historia UC](https://historia.uc.cl/) and IMFD research assistant, and new IMFD collaborating researchers, [Domingo Mery](https://www.ing.uc.cl/academicos-e-investigadores/domingo-arturo-mery-quiroz/), faculty member at [DCC UC](https://dcc.ing.uc.cl/); and [Hugo Rojas](https://derecho.uahurtado.cl/academicos/hugo-rojas/), director of [Cs. del Derecho en la U. Alberto Hurtado](https://derecho.uahurtado.cl/) and expert in transitional justice.

Driven by a task that is almost impossible to accomplish manually: **consolidating and combing through thousands of scanned documents, old photographs, and audio recordings scattered across various archives and collections**, [Nuestra MemorIA](https://nuestramemoria.ing.uc.cl/) was born as a transdisciplinary project aimed at developing innovative techniques specifically designed for the analysis of historical documents from the dictatorship. In this process, **the experience and research methods of experts in human rights, the humanities, and the social sciences are being incorporated.**
The ultimate goal is to **integrate fragmented information to help reconstruct historical knowledge**, thereby supporting the work of historians and social scientists.
[Nuestra MemorIA](https://nuestramemoria.ing.uc.cl/) is supported by the [Museo de la Memoria y los Derechos Humanos](https://mmdh.cl/), the [Instituto Nacional de Derechos Humanos (INDH)](https://www.indh.cl/), the [Vicaría de Solidaridad](https://www.vicariadelasolidaridad.cl/), the [Comité para la Prevención de la Tortura](https://mnpt.cl/), and the [Subsecretaría de Derechos Humanos](https://www.minjusticia.gob.cl/subsecretaria-de-derechos-humanos-2/). The project shares updates on its progress through the [Nuestra MemorIA](https://link.chtbl.com/MemorIA) podcast, available on all listening platforms, and on Instagram at [instagram.com/nuestramemoria.cl](http://instagram.com/nuestramemoria.cl) and on X (formerly Twitter) at [x.com/nuestramemo](http://x.com/nuestramemo).
##### **Nuestra MemorIA Seminar**
The initiative was presented on Tuesday, June 25, at a seminar attended by [Pedro Bouchon](https://www.ing.uc.cl/academicos-e-investigadores/pedro-alejandro-bouchon-aguirre/), Vice President for [Research at UC](https://www.uc.cl/investigacion/); [Loreto Valenzuela](https://ingenieriabiologicaymedica.uc.cl/es/personas/academicos/69-loreto-valenzuela), Dean of [Ingeniería UC](https://www.ing.uc.cl/); [Patricio Bernedo](https://historia.uc.cl/planta-academica/planta-ordinaria-especial-docente/56-bernedo-patricio), director of the [Centro UC para el Diálogo y la Paz](https://dialogoypaz.uc.cl/); [Felipe Mallea](https://coes.cl/integrantes/felipe-mallea/), Head of the Studies Department at the [Servicio Nacional de Migraciones](https://serviciomigraciones.cl/), the team behind the project, and nearly 100 attendees.
The research team faces several challenges. First, although the information exists, it is scattered across various organizations that have worked on its preservation. Then there is the sheer volume: **the Vicaría de la Solidaridad alone holds more than 85,000 records, a figure that nearly triples at the Museo de la Memoria**. The Judiciary holds more than 10 million pages, in addition to material gathered by various commissions, the Archivo Nacional, the Servicio Nacional de Migraciones, universities, and other organizations.
“What do we do with everything that is scattered, stored across so many institutions?” asked Hugo Rojas, who is also a researcher at the [Millennium Institute VioDemos](https://viodemos.cl/es/). “This cannot be lost, it must endure for future generations,” Rojas added: these archives are essential to transitional justice processes, “to access the truth, so that this never happens again.”
Collaboration will be key to taking on this task, said Pedro Bouchon, who highlighted that **the agreements in place with the Museo de la Memoria and the Vicaría de la Solidaridad make it possible to put science and engineering at the service of society**. For Dean Loreto Valenzuela, the academic community must contribute to the search for and recovery of memory, using tools that today **”can allow us to discover information that would otherwise remain hidden.”**
##### **A photo, a face, a name**
Another challenge posed by working with **millions of files from the dictatorship is that they consist of documents in the most diverse formats**: handwritten and typewritten texts, photos, videos and audio recordings, letters and drawings—all containing scattered data on victims and witnesses of human rights violations.
Domingo Mery described how they have begun to **test the effectiveness of different computational and artificial intelligence tools**, such as Gemini or ChatGPT, in recovering information. In the field of visual recognition, he explained, AI systems have been able to recognize images, group them with others of similar characteristics, and describe them in text form. In addition, they are studying image quality enhancement, techniques for facial detection, and facial recognition across other images. In one test, he noted, using a black-and-white photo of a person taken before they were forcibly disappeared, the AI model found them in another image: on a sign carried by a family member during a march for the disappeared detainees.
Techniques have also been applied to **convert handwriting or old audio recordings into digital text**, from which a data repository can then be generated to begin further work.
##### **The power of machine learning**
Having the historical archive data in text format is certainly a first step. **But reviewing it individually is a task that is almost impossible to do manually**. This is where the most advanced **natural language processing (NLP)** techniques come into play, said Jocelyn Dunstan Escudero.
The first goal, the researcher noted, would be to generate an **annotated corpus in Spanish: a repository of concepts, such as names or places, reviewed, systematized, and labeled**. This is achievable: the researcher’s work has already produced the first annotated corpus of primary health care in Spanish, which took about two years to develop, a period during which around 10,000 annotations were generated. With this corpus, it was possible to study waiting lists in the country.
To achieve this goal in the field of human rights, Jocelyn Dunstan noted, **the support of social scientists—linguists, historians, political scientists, among many others—is required** to label and classify the digitized data. An annotated corpus will make it possible to train computers to carry out this same task automatically and on massive volumes of files, quickly recognizing entities such as cities, dates, or the names of captors.
##### **Networks, connections, relationships**
With information such as that described, classified and labeled, it will be possible to build graph databases, **systems that allow data to be stored with many more and different characteristics**. In a graph, a piece of data such as a person’s name can be linked to a date, witnesses, and a location—the last place they were seen alive. **Thanks to these additional features, graph databases function as networks that connect information more quickly, explained Juan Reutter**, enabling more targeted searches.
What is needed to build a graph database? Data. Thousands of pieces of data, like those that Jocelyn Dunstan’s research line hopes to systematize. Only in this way will it be possible to build that immense network in which each connection or node, linked to others, allows concepts to be explored, with the hope of detecting relationships that might otherwise have gone unnoticed and shedding light where shadows still persist.
