Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces semCER, a severity-weighted Character Error Rate for OCR and VLMs, by assigning edit costs based on an LLM judge's assessment of grapheme severity, without requiring word segmentation. semCER aligns with human reader ratings across multiple languages and demonstrates superior agreement with human judgment compared to normalized CER when discrepancies arise. The proposed method effectively differentiates the impact of various errors, such as minor diacritic mistakes versus critical letter substitutions, offering a more nuanced evaluation of transcription quality.
Standard OCR metrics treat a typo and a wrong name equally, but LLM judges reveal severity differences that humans care about, leading to better evaluation.
Character Error Rate (CER), the standard metric for OCR and vision-language models (VLMs), charges a typo as much as a wrong name. We introduce semCER, which keeps CER's alignment and denominator and charges each grapheme edit a severity from an LLM judge, with no word segmentation. Normalized CER keeps the score of a whole transcription, where it tracks readers as well as semCER or better, and semCER breaks that score down into named edits. The charges track readers'ratings edit by edit in English, Vietnamese, and Chinese and put the edits readers rate most severe near the top of a sorted list. On pairs drawn where CER and semCER disagreed, readers side with semCER more often than with normalized CER in English and Chinese, and equally often in Vietnamese. Across 16 languages, the judge charges a wrong diacritic or tone mark almost nothing in Spanish but nearly a wrong letter's severity in Telugu. Where we can count real words, a general VLM's wrong letter forms one far above chance while a classical pipeline's stays near it. We release the 61 million judged edits, the aggregated reader ratings, and the distilled scorer at https://github.com/nguyenx2t/semcer.