News from MLA2S and Make it (net)work!

• Claudius Krause (HEPHY Vienna (ÖAW))
• Nicki Holighaus (Acoustics Research Institute, Austrian Academy of Sciences)
• Claus Trost (Erich Schmid Institute of Materials Science of theAustrian Academy of Sciences)
• Kati Heinrich (IGF | ÖAW)
• Jan Odstrčilík (IMAFO)

Including a short round table news update.

Introduction to Research at ACDH

• Alexandra Lenz (ACDH)

Using LLMs to write a historical dialect dictionary: challenges and opportunities for the WBÖ

• Daniel Elsner (ACDH)
• Wolfgang Koppensteiner (ACDH)
• Katharina Korecky-Kröll (ACDH)
• Philipp Stöckle (ACDH)

In this talk, we investigate the potential of LLMs in supporting lexicographic work on non[1]standard linguistic varieties using data from the Dictionary of Bavarian Dialects in Austria (WBÖ). Based on approx. 2.4 million digitized and TEI-encoded dialect paper slips published via the Lexical Information System Austria (LIÖ), we construct a domain-specific corpus and evaluate LLMs in semantic classification and dictionary entry generation. Key preparatory steps include metadata enrichment, glossary and ontology development, and prompt engineering combined with Retrieval-Augmented Generation (RAG) techniques. Preliminary results suggest that LLMs can assist in organizing dialectal material into coherent semantic groupings. However, challenges persist regarding data preprocessing, structural conformity, and selection of representative examples. We discuss methodological implications and outline future directions, including the integration of agent-based systems and fine-tuning approaches tailored to dialect resources. This study contributes to the broader discourse on AI-assisted lexicography, highlighting both the potential and limitations of current LLM technologies in handling underrepresented language varieties.

Discussion

Using Word Embeddings to analyse differences and shifts of semantics over time, space, and corpora: Overview of architectures, implementations, and evaluations, with a focus of applications on the Austria Media Corpus.

• Stefan Resch (ACDH)

Word Embeddings are at the core of the majority of today's NLP architectures. They allow for a mathematical approximation of semantics by learning the contextual patterns of words and by mapping those to high dimensional vector spaces. In that space, words which are semantically related to each other, also map to vectors that are equally close to each other. We utilize these architectures to detect differences and shifts of meanings of words by comparing their respective vectors across time, space, and corpora. In this talk, we will give an overview of the used architectures and their implementations: static word embeddings suchs as word2vec, GloVe, fastText, and contextual ones such as BERT and RoBERTa, and how they are evaluated and applied on the Austrian Media Corpus.

Discussion

Model Context Protocol (MCP) Servers for Research, Automation, and AI Experiments

• Robert Klugseder (ACDH)

The Model Context Protocol is an open standard introduced by Anthropic in November 2024 that enables Large Language Models to interact with external tools, APIs, and data sources in a structured manner. MCP servers function as a standardized interface and solve the problem of fragmented integrations: Instead of developing individual connections for each combination of AI application and data source, MCP enables a universal architecture. Hosts, clients, and servers communicate via JSON-RPC 2.0. The three core primitives – Tools, Resources, and Prompt Templates – enable modular, reusable solutions and foster a rapidly growing open-source ecosystem for scalable, context-aware AI systems. In my presentation, I will briefly explain the fundamentals of the protocol before presenting application examples from my scholarly practice and self-developed servers.

Discussion

Networking with refreshments / further discussions