Human evaluations lay at the heart of evaluations within the field of Natural Language Processing (NLP). Seen as the “golden standard” of evaluations, questions are being asked on whether these evaluations are both reproducible and repeatable. One overlooked aspect is the design choices made by researchers when designing user interfaces (UIs). In this paper, four UIs used in past NLP human evaluations are assessed by UX experts, based on standardized human-centered interaction principles. Building on these insights, we derive several recommendations that the NLP community should apply when designing UIs, to enable more consistent human evaluation responses.

We examine evaluation of faithfulness to input data in the context of hotel highlights—brief LLM-generated summaries that capture unique features of accommodations. Through human evaluation campaigns involving categorical error assessment and span-level annotation, we compare traditional metrics, trainable methods, and LLM-as-a-judge approaches. Our findings reveal that simpler metrics like word overlap correlate surprisingly well with human judgments (r=0.63), often outperforming more complex methods when applied to out-of-domain data. We further demonstrate that while LLMs can generate high-quality highlights, they prove unreliable for evaluation as they tend to severely under- or over-annotate. Our analysis of real-world business impacts shows incorrect and non-checkable information pose the greatest risks. We also highlight challenges in crowdsourced evaluations.

Hallucinations are one of the most pressing challenges for large language models (LLMs). While numerous methods have been proposed to detect and mitigate them automatically, human evaluation continues to serve as the gold standard. However, these human evaluations of hallucinations show substantial variation in definitions, terminology, and evaluation practices. In this paper, we survey 64 studies involving human evaluation of hallucination published between 2019 and 2024, to investigate how hallucinations are currently defined and assessed. Our analysis reveals a lack of consistency in definitions and exposes several concerning methodological shortcomings. Crucial details, such as evaluation guidelines, user interface design, inter-annotator agreement metrics, and annotator demographics, are frequently under-reported or omitted altogether.

In this paper we describe our attempt to reproduce a single human evaluation quality criterion of the human evaluation that was in conducted in the paper “NeuralREG: An end-to-end approach to referring expression generation”. In particular, this paper describes the approach and challenges involved in reproducing the human evaluation as done by the original authors of the paper, the results obtained, and what insights we have gained from attempting this particular reproduction. Insights that we hope will enable refinements to both how human evaluations are documented by author(s) and enable better reproductions of NLP experiments in the future.

We describe our implementation and evaluation of the Hotel Highlights system which has been deployed live by trivago. This system leverages a large language model (LLM) to generate a set of highlights from accommodation descriptions and reviews, enabling travellers to quickly understand its unique aspects. In this paper, we discuss our motivation for building this system and the human evaluation we conducted, comparing the generated highlights against the source input to assess the degree of hallucinations and/or contradictions present. Finally, we outline the lessons learned and the improvements needed.

Automatic metrics are extensively used to evaluate Natural Language Processing systems. However, there has been increasing focus on how the are used and reported by practitioners within the field. In this paper, we have conducted a survey on the use of automatic metrics, focusing particularly on natural language generation tasks. We inspect which metrics are used as well as why they are chosen and how their use is reported. Our findings from this survey reveal significant shortcomings, including inappropriate metric usage, lack of implementation details and missing correlations with human judgements. We conclude with recommendations that we believe authors should follow to enable more rigour within the field.

We present an overview of the GEM 2024 shared task, which comprised of both data-to-text generation and summarization. New datasets were compiled specifically for the task to reduce data contamination in the large language models, which the participants were likely to use. The paper describes the tasks, the datasets, the participating systems, the evaluation methods, and some preliminary results. The full results will be presented at INLG ‘24.

At the heart of the Pyramid evaluation method for text summarization lie human written summary content units (SCUs). These SCUs areconcise sentences that decompose a summary into small facts. Such SCUs can be used to judge the quality of a candidate summary, possibly partially automated via natural language inference (NLI) systems. Interestingly, with the aim to fully automate the Pyramid evaluation, Zhang and Bansal (2021) show that SCUs can be approximated by automatically generated semantic role triplets (STUs). However, several questions currently lack answers, in particular: i) Are there other ways of approximating SCUs that can offer advantages?ii) Under which conditions are SCUs (or their approximations) offering the most value? In this work, we examine two novel strategiesto approximate SCUs: generating SCU approximations from AMR meaning representations (SMUs) and from large language models (SGUs), respectively. We find that while STUs and SMUs are competitive, the best approximation quality is achieved by SGUs. We also show through a simple sentence-decomposition baseline (SSUs) that SCUs (and their approximations) offer the most value when rankingshort summaries, but may not help as much when ranking systems or longer summaries.

To prevent the costly and inefficient use of resources on low-quality annotations, we want a method for creating a pool of dependable annotators who can effectively complete difficult tasks, such as evaluating automatic summarization. Thus, we investigate the recruitment of high-quality Amazon Mechanical Turk workers via a two-step pipeline. We show that we can successfully filter out subpar workers before they carry out the evaluations and obtain high-agreement annotations with similar constraints on resources. Although our workers demonstrate a strong consensus among themselves and CloudResearch workers, their alignment with expert judgments on a subset of the data is not as expected and needs further training in correctness. This paper still serves as a best practice for the recruitment of qualified annotators in other challenging annotation tasks.

We describe in this paper an attempt to reproduce some of the human of evaluation results from the paper “It’s not Rocket Science: Interpreting Figurative Language in Narratives”. In particular, we describe the methodology used to reproduce the chosen human evaluation, the challenges faced, and the results that were gathered. We will also make some recommendations on the learnings obtained from this reproduction attempt and what improvements are needed to enable more robust reproductions of future NLP human evaluations.

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.

Earlier research has shown that few studies in Natural Language Generation (NLG) evaluate their system outputs using an error analysis, despite known limitations of automatic evaluation metrics and human ratings. This position paper takes the stance that error analyses should be encouraged, and discusses several ways to do so. This paper is based on our shared experience as authors as well as a survey we distributed as a means of public consultation. We provide an overview of existing barriers to carrying out error analyses, and propose changes to improve error reporting in the NLG literature.

Data augmentation is an important method for evaluating the robustness of and enhancing the diversity of training data for natural language processing (NLP) models. In this paper, we present NL-Augmenter, a new participatory Python-based natural language (NL) augmentation framework which supports the creation of transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of NL tasks annotated with noisy descriptive tags. The transformations incorporate noise, intentional and accidental human mistakes, socio-linguistic variation, semantically-valid style, syntax changes, as well as artificial constructs that are unambiguous to humans. We demonstrate the efficacy of NL-Augmenter by using its transformations to analyze the robustness of popular language models. We find different models to be differently challenged on different tasks, with quasi-systematic score decreases. The infrastructure, datacards, and robustness evaluation results are publicly available on GitHub for the benefit of researchers working on paraphrase generation, robustness analysis, and low-resource NLP.

Evaluations in machine learning rarely use the latest metrics, datasets, or human evaluation in favor of remaining compatible with prior work. The compatibility, often facilitated through leaderboards, thus leads to outdated but standardized evaluation practices. We pose that the standardization is taking place in the wrong spot. Evaluation infrastructure should enable researchers to use the latest methods and what should be standardized instead is how to incorporate these new evaluation advances. We introduce GEMv2, the new version of the Generation, Evaluation, and Metrics Benchmark which uses a modular infrastructure for dataset, model, and metric developers to benefit from each other’s work. GEMv2 supports 40 documented datasets in 51 languages, ongoing online evaluation for all datasets, and our interactive tools make it easier to add new datasets to the living benchmark.

We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for the 2021 shared task at the associated GEM Workshop.

Common sense is an integral part of human cognition which allows us to make sound decisions, communicate effectively with others and interpret situations and utterances. Endowing AI systems with commonsense knowledge capabilities will help us get closer to creating systems that exhibit human intelligence. Recent efforts in Natural Language Generation (NLG) have focused on incorporating commonsense knowledge through large-scale pre-trained language models or by incorporating external knowledge bases. Such systems exhibit reasoning capabilities without common sense being explicitly encoded in the training set. These systems require careful evaluation, as they incorporate additional resources during training which adds additional sources of errors. Additionally, human evaluation of such systems can have significant variation, making it impossible to compare different systems and define baselines. This paper aims to demystify human evaluations of commonsense-enhanced NLG systems by proposing the Commonsense Evaluation Card (CEC), a set of recommendations for evaluation reporting of commonsense-enhanced NLG systems, underpinned by an extensive analysis of human evaluations reported in the recent literature.

We observe a severe under-reporting of the different kinds of errors that Natural Language Generation systems make. This is a problem, because mistakes are an important indicator of where systems should still be improved. If authors only report overall performance metrics, the research community is left in the dark about the specific weaknesses that are exhibited by ‘state-of-the-art’ research. Next to quantifying the extent of error under-reporting, this position paper provides recommendations for error identification, analysis and reporting.

This paper describes an attempt to reproduce an earlier experiment, previously conducted by the author, that compares hedged and non-hedged NLG texts as part of the ReproGen shared challenge. This reproduction effort was only able to partially replicate results from the original study. The analyisis from this reproduction effort suggests that whilst it is possible to replicate the procedural aspects of a previous study, replicating the results can prove more challenging as differences in participant type can have a potential impact.

Human assessment remains the most trusted form of evaluation in NLG, but highly diverse approaches and a proliferation of different quality criteria used by researchers make it difficult to compare results and draw conclusions across papers, with adverse implications for meta-evaluation and reproducibility. In this paper, we present (i) our dataset of 165 NLG papers with human evaluations, (ii) the annotation scheme we developed to label the papers for different aspects of evaluations, (iii) quantitative analyses of the annotations, and (iv) a set of recommendations for improving standards in evaluation reporting. We use the annotations as a basis for examining information included in evaluation reports, and levels of consistency in approaches, experimental design and terminology, focusing in particular on the 200+ different terms that have been used for evaluated aspects of quality. We conclude that due to a pervasive lack of clarity in reports and extreme diversity in approaches, human evaluation in NLG presents as extremely confused in 2020, and that the field is in urgent need of standard methods and terminology.

This paper describes the implementation of the Hotel Scribe system. A commercial Natural Language Generation (NLG) system which generates descriptions of hotels from accommodation metadata with a high level of content and linguistic variation in English. It has been deployed live by *Anonymised Company Name* for the purpose of improving coverage of accommodation descriptions and for Search Engine Optimisation (SEO). In this paper, we describe the motivation for building this system, the challenges faced when dealing with limited metadata, and the implementation used to generate the highly variate accommodation descriptions. Additionally, we evaluate the uniqueness of the texts generated by our system against comparable human written accommodation description texts.