ADAGIO – Adaptive Text Generation with Generative AI

ADAGIO

ADAGIO – Adaptive Text Generation with Generative AI

Technologies:

Language Processing

Related Solutions:

Language and Voice Technologies

Aim of the project

The main goal of ADAGIO is the development of a text generation system which can be adapted to specific use cases and domains. The project investigates the use of Generative Artificial Intelligence methods and models, in particular Large Language Model. Adaptation, applied to real-world use cases in the industry.
 

Research and development 

ADAGIO centres on use cases in Basque, Spanish and English, within the Generative AI paradigm, and explores the adaptation of Large Language Models to assist users in their respective tasks. Several approaches are being investigated within the project, mainly variants of zero-shot and few-shot prompting, along with efficient fine-tuning and retrieval- augmented generation. The project also dedicates part of its activities to the design and development of neural optimisation methods and models, to achieve responsive systems that could be deployed in computational environments with limited resources.

 

Technologies

  • Large Language Models
  • Prompting & fine-tuning
  • Retrieval Augmented Generation

 

Field of application

The project tackles several sectors and fields of application, based on the needs and requirements of the companies in its consortium. It addresses in particular content generation in several industrial domains, health and physical activity, games and marketing.

 

Project partners

 

Project financing

  • Hazitek Programme of the SPRI group (2023)

Looking for support for your next project? Contact us, we are looking forward to helping you.

Vicomtech

Parque Científico y Tecnológico de Gipuzkoa,
Paseo Mikeletegi 57,
20009 Donostia / San Sebastián (Spain)

+(34) 943 309 230

Zorrotzaurreko Erribera 2, Deusto,
48014 Bilbao (Spain)

close overlay

Behavioral advertising cookies are necessary to load this content

Accept behavioral advertising cookies