As a professional who works with the power and complexity of words on a daily basis, I know that language can be a weapon of both truth and manipulation, which underscores the value of precise detection tools.
GATE Institute‘s hands-on workshop on “Intelligent Government & Disinformation – Tools, Resources, and Trends in Elections” presented a compelling trifecta of topics: detecting disinformation, understanding its electoral implications, and monitoring its regulatory frameworks. These subjects of exploration are not just academic curiosities; they strike at the heart of our democratic principles.
The sessions opened with a study of the documents of the Bulgarian National Assembly, aiming to determine, among other things, how useful the institution is as a data source and how sensitive topic models are to this kind of open data.
Next, I learned about the “emotional IQ” of AI, if I can put it that way – a groundbreaking automatic emotion detection tool for short Bulgarian sentences. Launched in 2022 as the first of its kind, the tool allows users to enter a sentence and instantly receive a classification of its emotional tone. It identifies emotions across 11 categories, such as joy, fear, and anger.
HurtLex was yet another engaging concept I wanted to learn more about. It is a comprehensive lexicon of hate words that can be utilized to detect and analyze hate speech in social media from a multilingual perspective. Its foundation lies in the Italian hate lexicon, which organizes harmful terms into 17 semantic groups. On a side note, the notion of having 17 distinct categories for “words that hurt” really made me pause. It’s an unsettling but essential way of showing how hate speech can take on many different forms and how it can be tracked across various languages.
I was clearly fascinated to meet Professor Kiril Simov, a leading European scientist and researcher in the field of computational linguistics, representing the Artificial Intelligence and Language Technologies Department, IICT, Bulgarian Academy of Sciences.
The Disinformation Trends and Elections in 2024 panel took a sobering turn while speakers explored various forms of disinformation tactics.
#HandsOffMyHighSchool left a strong impression on the audience*. It was obvious how deeply the science behind the topic resonated with the two presenters.
They highlighted how even accurate information, when stripped of context, can become vile propaganda.
Here, I thought about translation, which has always been and always will be about context. I thought about the infamous case of mistranslation during the Cold War, when a poorly contextualized phrase nearly escalated into a nuclear threat (see paragraph Keep digging).
From an idealistic perspective, I left the event realizing that it’s not just about the “words” themselves – it’s about their deeper meaning, their impact, and their connection to truth. Yet, in a more practical sense, words are sometimes precisely what they are – units of data, naturally processed into “the bigger picture”.
*The case study is free to download on A Data Pro’s website https://lnkd.in/e673Wxm9