🤖 AI Summary
This study addresses the cumbersome and resource-intensive manual review process for collateral eligibility in securities prospectuses, which is challenged by lengthy, semi-structured documents often mixing German and English. The work proposes a generative information extraction framework—novel in its application of large language models (LLMs) to central bank securities eligibility assessment—that decomposes the task into extraction, standardization, and explanation stages to effectively handle OCR noise and multilingual interference. It introduces an innovative value-based LLM-as-a-judge evaluation approach, overcoming limitations of conventional position-based metrics. The system achieves 91% precision at the document-level eligibility decision, demonstrating conservative and reliable performance with significantly reduced risk of erroneously accepting ineligible securities.
📝 Abstract
Verifying the eligibility of securities as collateral is a key responsibility of the German Central Bank. However, manually verifying these assets against legal and financial criteria within lengthy, semi-structured, and often bilingual prospectuses is a resource-intensive task. While previous efforts utilized traditional Named Entity Recognition (NER) for information extraction, these methods can struggle with OCR noise, linguistic variance, and rigid span-based constraints, and the need for manually annotated training data for each relevant annotation type. In this paper, we present the first case study applying Large Language Models (LLMs) to the eligibility examination process, shifting the paradigm toward a generative Information Extraction pipeline. Our approach decomposes the task into extraction, normalization, and interpretation, allowing for greater flexibility in handling noisy text and interleaved German-English content. We further introduce a value-based evaluation methodology using LLM-as-a-judge, which offers a more semantic assessment than location-based metrics. Our results demonstrate that LLM-based systems achieve high precision (up to 91%) in document-level eligibility, exhibiting a conservative operating profile that minimizes false acceptance.