In order to diagnostically analyze and improve the capability of pretrained language models (PLMs) in text generation, we propose TGEA 2.0, to date the largest dataset built on machine-authored texts by PLMs with fine-grained semantic annotations on a wide variety of pathological generation errors. We collect 170K nominal, phrasal and sentential prompts from 6M natural sentences in 3 domains. These prompts are fed into 4 generative PLMs with their best decoding strategy to generate paragraphs.

Table 2: The number of erroneous texts generated with different decoding strategies. Figure 2: The distribution of MiSEW over the number of tokens contained in each MiSEW . We have fine-tuned several commonly used Chinese PLMs as baselines. All models have 12 attention heads and the hidden size is 768. We train these models on 8 Tesla P100 with 16G memory.

In order to diagnostically analyze and improve the capability of pretrained language models (PLMs) in text generation, we propose TGEA 2.0, to date the largest dataset built on machine-authored texts by PLMs with fine-grained semantic annotations on a wide variety of pathological generation errors. We collect 170K nominal, phrasal and sentential prompts from 6M natural sentences in 3 domains. These prompts are fed into 4 generative PLMs with their best decoding strategy to generate paragraphs. We define a \textbf{Mi}nimal \textbf{S}et of \textbf{E}rror-related \textbf{W}ords (MiSEW) for each erroneous span, which not only provides error-associated words but also rationalizes the reasoning behind the error. Quality control with a pre-annotation and feedback loop is performed before and during the entire annotation process.