Umm... With Transformers? Insights from Filled Pause Use across Four Slavic Parliaments

2026-07-07 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Social Sciences & Behavioral Studies, Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, medium

Summary

A study analyzed approximately 4,000 hours of parliamentary speech from four Slavic languages—Croatian, Czech, Polish, and Serbian—to investigate the use of filled pauses (FPs). Researchers employed transformer-based automatic detection for FPs and Generalised Estimating Equations (GEE) with Mundlak correction to model FP rates, distinguishing between within- and between-speaker effects. The analysis confirmed a negative correlation between age and speech rate with FP rate. Interestingly, gender effects on FP rates were found to be language-specific and contrary to most existing literature. Novel findings also revealed a consistent positive association between speaker sentiment and FP rate. Furthermore, political orientation and power status modulated FP rates in a parliament-specific manner, with opposition speakers generally exhibiting lower FP rates compared to those in governing coalitions.

Key takeaway

For AI Scientists developing speech analysis models for political discourse, you should account for language-specific gender effects and the influence of speaker sentiment and political affiliation on filled pause rates. Your models should integrate transformer-based detection and GEE for robust analysis across diverse linguistic and political contexts. This approach ensures more accurate and contextually aware interpretations of spontaneous speech patterns.

Key insights

Transformer-based detection reveals nuanced filled pause patterns in Slavic parliamentary speech, challenging prior assumptions on gender and linking FPs to sentiment and political status.

Principles

• Age and speech rate negatively correlate with filled pause rates.
• Gender effects on filled pauses are language-specific.
• Positive sentiment associates with higher filled pause rates.

Method

Filled pause occurrence is detected using transformer-based models. FP rate is modeled via Generalised Estimating Equations (GEE) with Mundlak correction to separate within- and between-speaker effects.

In practice

• Apply transformer models for large-scale speech analysis.
• Use GEE with Mundlak correction for speaker-level effects.
• Consider sentiment and political context in speech analysis.

Topics

• Filled Pauses
• Transformer Models
• Slavic Languages
• Parliamentary Speech
• Speech Analysis
• Sociolinguistics

Best for: AI Scientist, Research Scientist

Related on AIssential

• Mapping the Political Discourse in the Brazilian Chamber of Deputies: A Multi-Faceted Computational Approach — Computational methods can reveal hidden patterns in parliamentary discourse beyond voting records.
• The Dynamics of Human and AI-Generated Language: How Semantics Fluctuates across Different Timescales — The ACW-0 semantic timescale analysis effectively distinguishes temporal organization of generic versus specific content in human and AI speech.
• Beyond Acoustic Emotion Recognition: Multimodal Pathos Analysis in Political Speech Using LLM-Based and Acoustic Emotion Models — LLM-based multimodal analysis captures political emotion better than acoustic models, despite acoustic utility for Arousal.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• Analysing Differences in Persuasive Language in LLM-Generated Text: Uncovering Stereotypical Gender Patterns — LLMs exhibit pervasive gender-stereotypical persuasive language biases across models and languages, reflecting societal norms.
• Understanding Sarcasm in Generative AI Models: How ChatGPT Interprets Human Intent — Sarcasm detection in AI requires understanding non-literal meaning through contextual learning, moving beyond simple keyword analysis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.