LBRAND, a tool capable of automatically analysing the social brand of Basque and Catalan:

Data monitoring and analytics

Overview

In the LBRAND project we developed an artificial intelligence tool capable of automatically and continuously analysing the social brand of Basque and Catalan. The system collects news stories from the digital press and messages from social networks, identifies texts about these languages and evaluates the social perception of these brands using a set of indicators. The tool is designed for the EHU’s research group NIK, and its function is to analyse the Basque and Catalan brands on a large scale.

Challenge

The main challenge consisted in automating the handling of large text corpora from several years and different sources —news stories from 37 digital media and messages from the X platform— and identifying those that are actually about Basque or Catalan. As well as evaluating the presence of these languages, the project sought to reliably infer aspects such as the attitudes revealed by the texts, the entities and people associated with these languages, the main topics and the location of each brand —with the themes rural environment versus urban environment and tradition versus modernity—.

Collaboration

It was developed in collaboration with the EHU’s research group NIK.

Result

The result is the interactive platform LBRAND, which allows consultation of the main indicators used for analysing the Basque and Catalan brands: their level of presence, attitudes towards the brands, and personalities associated with the languages, the thematic associations and symbolic locations. To calculate these indicators automatically, the system uses advanced Natural Language Understanding (NLU) techniques based on the LLMs RoBERTa and Llama 3.3 70B, among others. It is also capable of analysing news articles from the period 2008-2024 and social media messages from the period 2008-2022.