Tgea: An error-annotated dataset and benchmark tasks for text generation from pretrained language models

TL;DR

TGEA introduces the first error-annotated dataset for PLM-generated text, providing a comprehensive 24-type error taxonomy, associated spans, minimal corrections, and rationales to enable diagnostic evaluation. It pairs this dataset with five benchmark tasks—errant text detection, erroneous/associated span detection, error type classification, error correction, and rationale generation—and reports baselines using Chinese PLMs, revealing substantial room for progress. The dataset, sourced from a Chinese GPT-2 lineage (NEZHA-Gen), highlights substantial commonsense and discourse-related errors in machine-generated text, underscoring the gap between human and machine generation. By enabling interpretable error analysis and potential automatic correction, TGEA offers a valuable resource for advancing diagnostic and corrective methods in PLM-based text generation.

Abstract

In order to deeply understand the capability of pretrained language models in text generation and conduct a diagnostic evaluation, we propose TGEA, an error-annotated dataset with multiple benchmark tasks for text generation from pretrained language models (PLMs). We use carefully selected prompt words to guide GPT-2 to generate candidate sentences, from which we select 47K for error annotation. Crowdsourced workers manually check each of these sentences and detect 12k erroneous sentences. We create an error taxonomy to cover 24 types of errors occurring in these erroneous sentences according to the nature of errors with respect to linguistics and knowledge (eg, common sense). For each erroneous span in PLM-generated sentences, we also detect another span that is closely associated with it. Each error is hence manually labeled with comprehensive annotations, including the span of the error, the associated span, minimal correction to the error, the type of the error, and rationale behind the error. Apart from the fully annotated dataset, we also present a detailed description of the data collection procedure, statistics and analysis of the dataset. This is the first dataset with comprehensive annotations for PLM-generated texts, which facilitates the diagnostic evaluation of PLM-based text generation. Furthermore, we use TGEA as a benchmark dataset and propose a series of automatic diagnosis tasks, including error detection, error type classification, associated span detection, error rationale generation, to further promote future study on the automatic error detection and correction on texts generated by pretrained language models.

Figures (2)

• Figure 1: The different stages of the annotation process for each machine-generated text according to the prompt in TGEA. Better viewed in color.
• Figure 2: Distribution over the level-1 and level-2 error types in TGEA.