Artificial intelligence as a support tool for corpus query formulation

Santrauka

The importance of corpus work and corpus-linguistic methods is steadily increasing in linguistic research and in the field of German as a Foreign Language (GFL), as they enable a comprehensive and systematic analysis of linguistic phenomena. Researchers often work with a large number of corpora that differ in their datasets and structures. One challenge in this work, however, is the diversity of query systems used across individual corpora. These systems can be complex and heterogeneous, which often requires a considerable amount of time to understand their functionality and use them effectively. This diversity can make the search for specific linguistic patterns or phenomena difficult and frustrating for both experienced users and beginners, as a great deal of time must be spent familiarizing oneself with the particularities of individual query systems. In response to these challenges, the integration of AI as a tool for formulating search queries in corpora may play a decisive role in the future.
In this context, the present empirical contribution examines the potential of AI-supported systems for formulating complex search queries in COSMAS II. The study draws on 50 typical corpus-linguistic search tasks of different types, formulated by inexperienced users without prior knowledge of computational linguistics. The AI-generated queries are analyzed with regard to their formal correctness, precision, and functionality.
Methodologically, the study is based on a qualitative analysis of the query results. The results indicate the conditions under which AI can be used as a supportive tool in corpus querying and where its limitations lie. The contribution also discusses what information needs to be available to the AI in order to effectively support corpus-linguistic search processes.