Beraz, automatic News Simplification in Basque to Improve Access to Information

The Challenge

People learning Basque, as well as those with reading difficulties, may struggle to understand some news articles because of their vocabulary or complex linguistic structures.

Simplifying this content requires adapting the language without changing the original meaning, facts or statements. This is particularly important when using artificial intelligence, as models can introduce interpretations or hallucinations that are not present in the original text.

To address this challenge, TOKIKOM and Vicomtech set out to develop a solution that would make local news in Basque easier to understand without compromising journalistic accuracy.

The Proposed Solution

After more than a year of joint work, Vicomtech and TOKIKOM have developed a system that combines artificial intelligence and natural language processing. The tool analyses each news article, identifies its main ideas and generates an initial version in simpler, more accessible Basque.

Using this extensive collection of simplified news articles, the team developed an automatic text simplification system that converts complex texts into simplified versions. This approach achieved the best results in both automated and human evaluations.

One of the main challenges has been ensuring that simplified versions preserve the original meaning without adding information or introducing new interpretations. Particular attention has also been paid to statements and direct quotations, which must be reproduced faithfully.

Simplified versions are clearly labelled by the media outlets to distinguish them from the original articles. In this way, the tool complements journalists' work rather than replacing it.

The Result

The solution is already available through several outlets on the TOKIKOM platform, including Goiena, ALEA, GUKA, Aiaraldea, Txintxarri and ERRAN.

The tool makes local news published in Basque more accessible to people learning the language and those with reading difficulties. It also enables linguistic simplification to be incorporated into the news publishing process.

The project demonstrates the potential of artificial intelligence to improve access to information and develop new tools for Basque language processing.