Currently they are not. However, we are actively engaged in identifying potential shortcomings of these models ¹. The goal is to enhance its specialization for newspapers interpretation tasks through fine-tuning.

1 - The models may underperform or return limited answers in some contexts (see "ChatGPT vs Gemini vs LLaMA on Multilingual Sentiment Analysis"). Continuous fine-tuning would improve the experience.

Upon submission of the URL, the model is queried only on the text of the article, thereby the URL is masked. This protocol ensures that the model does not incorporate any information pertaining to the newspapers, which could potentially introduce bias into the evaluation process. In essence, each assessment of the article remains independent and is grounded solely on the information contained within the article itself.

There are three reasons for encountering an error message:
1) Submission of an empty URL.
2) Submission of a malformed URL. A quick check involves entering the website into your browser to confirm its accessibility.
3) The URL is accurate, but NAV AI might be unable to process it. This may occur if the URL is not from a newspaper or if the article is behind a paywall, a firewall or is protected by a password. The article cannot be processed further.
It may happen that an URL is unprocessed for unknown reasons: each available model exhibits a degree of randomness and may occasionally not deliver the expected responses. In such cases, please try again by inserting the same URL.

You can freely reuse the results obtained by using NAV AI, including snapshots of our compass. If you do so, please reference our website as your source.

We employ one of the most commonly used models to depict the political spectrum ², incorporating a horizontal axis for socio-economic factors and a vertical axis for socio-cultural considerations.

2 - Heywood, Andrew (2017). Political Ideologies: An Introduction (6th ed.). Basingstoke: Macmillan International Higher Education.

While navigating the website, users are not asked to enter any piece of information except for the desired (public) URL to scan. No further meta-information (e.g. IP address) is ever stored in our servers or forwarded to OpenAI.

NAV AI is an agnostic tool, not affiliated to other associations or institutions nor endorsing any social, political or economical view. It efficiently links users' prompts and commercial AI to provide services. Any result obtained should therefore keep in mind the potential limitations of the selected Large Language Models, and any conclusion reached thereafter is the sole view and responsibility of the user.

Absolutely! If you're a journalist, political scientist, or tech enthusiast, I'd be delighted to discuss potential collaborations with you.