The evaluation of natural language generation (NLG) tasks is a significant and longstanding research area. With the recent emergence of powerful large language models (LLMs), some studies have turned to LLM-based automatic evaluation methods, which demonstrate great potential to become a new evaluation paradigm following traditional string-based and model-based metrics. However, despite the improved performance of existing methods, they still possess some deficiencies, such as dependency on references and limited evaluation flexibility. Therefore, in this paper, we meticulously construct a large-scale NLG evaluation corpus **NLG-Eval** with annotations from both human and GPT-4 to alleviate the lack of relevant data in this field. Furthermore, we propose **Themis**, an LLM dedicated to NLG evaluation, which has been trained with our designed multi-perspective consistency verification and rating-oriented preference alignment methods. Themis can conduct flexible and interpretable evaluations without references, and it exhibits superior evaluation performance on various NLG tasks, simultaneously generalizing well to unseen tasks and surpassing other evaluation models, including GPT-4.

We present a reproduction study of the human evaluation of the coverage of fact checking explanations conducted by Atanasova et al. (2020), as a team in Track B of ReproNLP 2024. The setup of our reproduction study is almost the same as the original study, with some necessary modifications to the evaluation guideline and annotation interface. Our reproduction achieves a higher IAA of 0.20 compared to the original study’s 0.12, but discovers a mismatch between the IAA calculated by us with the raw annotation in the original study and the IAA reported in the original paper. Additionally, our reproduction results on the ranks of three types of explanations are drastically different from the original experiment, rendering that one important conclusion in the original paper cannot be confirmed at all. The case study illustrates that the annotators in the reproduction study may understand the quality criterion differently from the annotators in the original study.

Some prior work has shown that LLMs perform well in NLG evaluation for different tasks. However, we discover that LLMs seem to confuse different evaluation criteria, which reduces their reliability. For further verification, we first consider avoiding issues of inconsistent conceptualization and vague expression in existing NLG quality criteria themselves. So we summarize a clear hierarchical classification system for 11 common aspects with corresponding different criteria from previous studies involved. Inspired by behavioral testing, we elaborately design 18 types of aspect-targeted perturbation attacks for fine-grained analysis of the evaluation behaviors of different LLMs. We also conduct human annotations beyond the guidance of the classification system to validate the impact of the perturbations. Our experimental results reveal confusion issues inherent in LLMs, as well as other noteworthy phenomena, and necessitate further research and improvements for LLM-based evaluation.

Research on automated text summarization typically uses human and automatic evaluation methods. While most recent studies focus on intrinsic evaluation, which assesses the general quality of summaries, e.g. coherence and informativeness, we concentrate on task-based extrinsic evaluation to determine the usefulness of summaries. We incorporate three downstream tasks, namely question answering, text classification, and text similarity assessment, and measure the usefulness of summaries for these tasks by several metrics. Our findings reveal that summaries are generally useful in tasks that require a comprehensive grasp of the text but are less useful in tasks requiring a more specific understanding of the text. We also analyze the usefulness and inherent properties of summaries from different models, and find that fine-tuned models consistently produce more useful summaries across all three tasks. In contrast, zero-shot models tend to lean towards text classification and similarity assessment, providing more general and less detailed summaries. Additionally, we assess the correlation between 14 intrinsic automatic metrics and human judgments. Intrinsic metrics perform well in evaluating summaries for question answering but are less effective in the other two tasks. This highlights the limitations of relying solely on intrinsic metrics for assessing summary performance and usefulness.

Factuality is important to dialogue summarization. Factual error correction (FEC) of model-generated summaries is one way to improve factuality. Current FEC evaluation that relies on factuality metrics is not reliable and detailed enough. To address this problem, we are the first to manually annotate a FEC dataset for dialogue summarization containing 4000 items and propose FERRANTI, a fine-grained evaluation framework based on reference correction that automatically evaluates the performance of FEC models on different error categories. Using this evaluation framework, we conduct sufficient experiments with FEC approaches under a variety of settings and find the best training modes and significant differences in the performance of the existing approaches on different factual error categories.

Cross-lingual summarization aims to help people efficiently grasp the core idea of the document written in a foreign language. Modern text summarization models generate highly fluent but often factually inconsistent outputs, which has received heightened attention in recent research. However, the factual consistency of cross-lingual summarization has not been investigated yet. In this paper, we propose a cross-lingual factuality dataset by collecting human annotations of reference summaries as well as generated summaries from models at both summary level and sentence level. Furthermore, we perform the fine-grained analysis and observe that over 50% of generated summaries and over 27% of reference summaries contain factual errors with characteristics different from monolingual summarization. Existing evaluation metrics for monolingual summarization require translation to evaluate the factuality of cross-lingual summarization and perform differently at different tasks and levels. Finally, we adapt the monolingual factuality metrics as an initial step towards the automatic evaluation of summarization factuality in cross-lingual settings. Our dataset and code are available at https://github.com/kite99520/Fact_CLS.

This paper reports a reproduction study of the human evaluation of role-oriented dialogue summarization models, as part of the ReproNLP Shared Task 2023 on Reproducibility of Evaluations in NLP. We outline the disparities between the original study’s experimental design and our reproduction study, along with the outcomes obtained. The inter-annotator agreement within the reproduction study is observed to be lower, measuring 0.40 as compared to the original study’s 0.48. Among the six conclusions drawn in the original study, four are validated in our reproduction study. We confirm the effectiveness of the proposed approach on the overall metric, albeit with slightly poorer relative performance compared to the original study. Furthermore, we raise an open-ended inquiry: how can subjective practices in the original study be identified and addressed when conducting reproduction studies?

We report our efforts in identifying a set of previous human evaluations in NLP that would be suitable for a coordinated study examining what makes human evaluations in NLP more/less reproducible. We present our results and findings, which include that just 13% of papers had (i) sufficiently low barriers to reproduction, and (ii) enough obtainable information, to be considered for reproduction, and that all but one of the experiments we selected for reproduction was discovered to have flaws that made the meaningfulness of conducting a reproduction questionable. As a result, we had to change our coordinated study design from a reproduce approach to a standardise-then-reproduce-twice approach. Our overall (negative) finding that the great majority of human evaluations in NLP is not repeatable and/or not reproducible and/or too flawed to justify reproduction, paints a dire picture, but presents an opportunity for a rethink about how to design and report human evaluations in NLP.

Dialogue summarization is receiving increasing attention from researchers due to its extraordinary difficulty and unique application value. We observe that current dialogue summarization models have flaws that may not be well exposed by frequently used metrics such as ROUGE. In our paper, we re-evaluate 18 categories of metrics in terms of four dimensions: coherence, consistency, fluency and relevance, as well as a unified human evaluation of various models for the first time. Some noteworthy trends which are different from the conventional summarization tasks are identified. We will release DialSummEval, a multi-faceted dataset of human judgments containing the outputs of 14 models on SAMSum.