Prof. Laurie Alice Vera Jiménez

Laurie Vera is an architect, researcher, and professor at the Faculty of Architecture, Design, and Arts of the National University of Asunción (FADA-UNA). She graduated from FADA-UNA and has a Master's degree in History, Civilizations, and Heritage, with specialization in "City, Architecture, and Heritage" (VAP) from Paris-Cité University and the National School of Architecture Paris-Vallée de la Seine (UPC-ENSAPVS). Besides, she holds a diploma in Latin American Studies and Master 1 in "Social Sciences of Latin America" from the Institute of Advanced Studies on Latin America at the Sorbonne Nouvelle (IHEAL).

Title: Can artificial intelligence understand a landscape?

Abstract

Integrating artificial intelligence into the field of cultural heritage opens up new possibilities for documenting, researching, interpreting, and managing assets, landscapes, and memories. However, it also raises a fundamental question: what heritage do we make visible when we convert it into data, and what knowledge, relationships, and voices might be left out of that process?

Drawing on research into the historic centers of San Lorenzo and Areguá in Paraguay, this presentation proposes shifting the focus from heritage understood as a collection of objects toward a relational conception—one in which architecture, nature, territory, memory, cultural practices, and communities form an inseparable web.

From this perspective, the presentation examines the potential of AI to connect scattered heritage information, document oral histories, analyze territorial transformations, and contribute to new forms of museum mediation. At the same time, it addresses challenges related to data bias, cultural representation, language, sovereignty over sensitive information, and community governance.

Ultimately, the proposal advocates for moving toward a heritage AI capable of contributing to the continuity of heritage without replacing the interpretive and decision-making capacities of the communities that sustain it.

Prof. Marvin Agüero Torales

Marvin Matías Agüero-Torales is a Paraguayan computer scientist and researcher specializing in Natural Language Processing (NLP), machine learning (ML), and Artificial Intelligence (AI) for low-resource languages. He holds a PhD in Information and Communication Technologies from the University of Granada, Spain, where his doctoral research focused on topic and sentiment analysis in multilingual opinions, including Guarani and Jopara. Precisely, his research addresses the development and evaluation of language technologies for Guarani and Jopara, spanning parallel corpora, semantic resources, or language models. He is involved in GuaranIA, an initiative promoting open AI resources and technologies for Paraguayan Guarani, and in collaborative efforts to strengthen Indigenous-language NLP, including work associated with the Lacuna Fund, the National University of Itapúa (UNI), and Universidad de la República (UdelaR). His research interests also encompass the AmericasNLP community and multilingual evaluation, including GUA-SPA or INCLUDE-2.0. His work seeks to advance more inclusive and culturally grounded AI systems that better reflect the linguistic diversity and knowledge of low-resource languages communities.

Title: From Language Resources to Cultural Knowledge: Building AI for Guarani and Beyond

Abstract

Language is more than a means of communication: it carries collective memory, cultural knowledge, and ways of understanding the world. Yet, Indigenous and other underrepresented languages remain poorly represented in many AI systems, raising important questions about linguistic equity, cultural representation, and whose knowledge becomes accessible in the digital age.

This talk explores the challenges and opportunities of developing AI for Guarani, drawing on research and practical experience across language resources, natural language processing, or multilingual evaluation. It begins with foundational work on Guarani and Guarani–Spanish (Jopara) language technologies, including parallel corpora, machine translation, semantic resources, or the analysis of code-switching in Jopara. It then discusses collaborative efforts to expand the digital resources available for Guarani, including initiatives associated with the Lacuna Fund, the National University of Itapúa (UNI), Universidad de la República (UdelaR), and GuaranIA, which seeks to foster the development of open language resources, models, and applications.

The presentation also examines broader efforts to advance Indigenous-language technologies through AmericasNLP, including research on machine translation, grammar generation or culturally grounded image captioning. In addition, it considers multilingual evaluation initiatives such as GUA-SPA or INCLUDE-2.0 and the challenge of determining whether AI systems can represent regional knowledge and cultural context, rather than simply generate linguistically plausible responses.

These experiences illustrate a central tension in the development of AI and cultural heritage: making a language available to AI does not necessarily mean that its cultural meanings, linguistic variation, and associated knowledge are adequately represented. Data scarcity, dialectal diversity, code-switching, and other limitations, drive the lack of culturally meaningful evaluation methods all influence what AI systems can learn and communicate.

The talk argues for an approach to inclusive and trustworthy AI that goes beyond collecting data or developing language models. Through the case of Guarani, it reflects on how AI can support cultural transmission while respecting the knowledge, agency, and diversity of the communities that sustain it.