Sounds interesting? Here's the specific research question:
Use of LLMs to Support Data Entry, Data Validation, and Data Quality Analysis
- In many optimization and analysis applications, the key bottleneck is not the modeling itself, but rather the structured and high-quality collection of input data.
- LLMs promise to assist in converting unstructured information into structured inputs and in automatically deriving data quality rules; however, their reliability has not yet been adequately evaluated.
- The aim is to examine at which stages of the data entry and data quality process LLMs provide robust added value and what control mechanisms are necessary to ensure this.
Here is how we envision this methodology:
- Development of a prototype that converts unstructured or semi-structured input into data structures relevant to OPTANO and identifies potential quality issues.
- Comparison of various approaches, such as pure extraction, rule-based post-validation, iterative self-checking, or human-in-the-loop.
- Evaluation based on accuracy, completeness, correction effort, processing time, and usefulness in practical workflows
You'll need the following knowledge:
- Strong knowledge of Python and an interest in data processing
- Basic knowledge of data engineering, data quality, or information extraction
- Interest in the experimental evaluation of LLM systems