Beyond Document Page Classification: Design, Datasets, and Challenges
This work identifies a critical problem for researchers and practitioners in document AI, as it is incremental in calling for updated benchmarks and methodologies.
The paper addresses the gap between document classification benchmarks and real-world applications by formalizing tasks and highlighting the lack of public multi-page datasets, showing that current benchmarks are irrelevant and need updates to evaluate complete documents as they occur in practice.
This paper highlights the need to bring document classification benchmarking closer to real-world applications, both in the nature of data tested ($X$: multi-channel, multi-paged, multi-industry; $Y$: class distributions and label set variety) and in classification tasks considered ($f$: multi-page document, page stream, and document bundle classification, ...). We identify the lack of public multi-page document classification datasets, formalize different classification tasks arising in application scenarios, and motivate the value of targeting efficient multi-page document representations. An experimental study on proposed multi-page document classification datasets demonstrates that current benchmarks have become irrelevant and need to be updated to evaluate complete documents, as they naturally occur in practice. This reality check also calls for more mature evaluation methodologies, covering calibration evaluation, inference complexity (time-memory), and a range of realistic distribution shifts (e.g., born-digital vs. scanning noise, shifting page order). Our study ends on a hopeful note by recommending concrete avenues for future improvements.}