Speech Technologies

rom automatic speech recognition (ASR) to large language models and multimodal, multilingual AI systems – this blog series explores current developments in language technologies. Our articles show how AI processes spoken and written language, and how it is used in areas like text simplification, auditing, or accessible parliamentary sessions.

Perfect for anyone curious about how language AI works today – and where it will be needed tomorrow.

00 Blog Fine tuning ASR - Lamarr Institute for Machine Learning (ML) and Artificial Intelligence (AI)

Basics

Fine-Tuning ASR Models: Enhancing Accuracy and Adaptability for Specific Tasks

30. April 2025

Fine-tuning improves pre-trained ASR models for specific tasks, adapting them to new languages, accents, and domains. This process enhances accuracy in the direction of interest, while requiring less computational cost than training models from scratch....

An open book with glowing binary code and light streams rising from its pages, symbolizing the fusion of storytelling and digital technology.

Research

Can Stories Help LLMs Reason? A Deep Dive into the Story of Thought (SoT) Framework

11. December 2024

Can storytelling unlock new reasoning powers in AI? Explore how the innovative Story of Thought (SoT) framework uses narratives to enhance LLMs' ability to solve complex, knowledge-intensive problems with greater coherence and depth....

Big data technology Data science analysing artificial intelligence generative AI deep learning machine learning algorithm Neural flow network analytics innovation abstract futuristic. 3d rendering

Research

Multilingual Instruction-Tuning: Do Polyglot Models Demand for Multilingual Instructions?

30. October 2024

Multilingual instruction-tuning of models pre-trained on Indo-European languages reveals that (semantically) parallel datasets significantly boost cross-lingual capabilities compared to monolingual ones. We introduce new multilingual evaluation resources and datasets for polyglot model training. ...

Conceptual image of a library filled with digital assets like images, documents, and videos.

Application

How Audio Mining, Generative AI, and LLMs Are Transforming Media Archive Searches

16. October 2024

Discover how Fraunhofer IAIS is revolutionizing media archive searches with Audio Mining, Generative AI, and Large Language Models, offering journalists and researchers smarter, more efficient ways to access valuable content....

Beitragsbild 5 scaled - Lamarr Institute for Machine Learning (ML) and Artificial Intelligence (AI)

Basics

Multimodality in Large Language Models – How AI is becoming more humanlike

19. June 2024

Large Language Models have demonstrated impressive capabilities in the last few years. Can we use Multimodality to make them even more powerful and make them think more humanlike? ...

Beitragsbild scaled - Lamarr Institute for Machine Learning (ML) and Artificial Intelligence (AI)

Application

Live-ASR Software Facilitates Barrier-Free Plenary Sessions in the Saxon State Parliament

22. May 2024

How can we ensure more inclusive politics? The Saxon State Parliament is an example of how a Live-ASR software developed by Fraunhofer IAIS helps to enable barrier-free plenary sessions....

Basics

Navigating the Evolution of Automatic Speech Recognition (ASR)

10. April 2024

Discover the evolution and challenges of Automatic Speech Recognition (ASR) technology, from statistical models to advanced neural networks. Delve into robustness, vocabulary limitations, and hallucinations, unveiling solutions for seamless Human Computer Interaction....

00 Blog Toborek Vanessa Forschung Sentence Allignment - Lamarr Institute for Machine Learning (ML) and Artificial Intelligence (AI)

Research

Simplified language for all: How to create a dataset in Simple German

06. December 2023

Machine Learning models are only as diverse as the datasets they are trained on. Learn more about the first publicly available bilingual dataset in German and Simple German language provided by us....

- Lamarr Institute for Machine Learning (ML) and Artificial Intelligence (AI)

Basics

Let us be clear: Why text simplification matters and what ML can do to help

15. November 2023

Text simplification is the key to written language accessibility. Discover what it entails, who benefits, and its link to readability assessment. Get an overview of ML approaches for automated text simplification....