We present version 1.3 of the PARSEME multilingual corpus annotated with verbal multiword expressions. Since the previous version, new languages have joined the undertaking of creating such a resource, some of the already existing corpora have been enriched with new annotated texts, while others have been enhanced in various ways. The PARSEME multilingual corpus represents 26 languages now. All monolingual corpora therein use Universal Dependencies v.2 tagset. They are (re-)split observing the PARSEME v.1.2 standard, which puts impact on unseen VMWEs. With the current iteration, the corpus release process has been detached from shared tasks; instead, a process for continuous improvement and systematic releases has been introduced.

There is a growing concern regarding the reproducibility of human evaluation studies in NLP. As part of the ReproHum campaign, we conducted a study to assess the reproducibility of a recent human evaluation study in NLP. Specifically, we attempted to reproduce a human evaluation of a novel approach to enhance Role-Oriented Dialogue Summarization by considering the influence of role interactions. Despite our best efforts to adhere to the reported setup, we were unable to reproduce the statistical results as presented in the original paper. While no contradictory evidence was found, our study raises questions about the validity of the reported statistical significance results, and/or the comprehensiveness with which the original study was reported. In this paper, we provide a comprehensive account of our reproduction study, detailing the methodologies employed, data collection, and analysis procedures. We discuss the implications of our findings for the broader issue of reproducibility in NLP research. Our findings serve as a cautionary reminder of the challenges in conducting reproducible human evaluations and prompt further discussions within the NLP community.

The WebNLG task consists of mapping a knowledge graph to a text verbalising the con- tent of that graph. The 2017 WebNLG edi- tion required participating systems to gener- ate English text from a set of DBpedia triples, while the 2020 WebNLG+ challenge addition- ally included generation into Russian and se- mantic parsing of English and Russian texts. In contrast, WebNLG 2023 focuses on four under-resourced languages which are severely under-represented in research on text genera- tion, namely Breton, Irish, Maltese and Welsh. In addition, WebNLG 2023 once again includes Russian. In this paper, we present the organi- sation of the shared task (data, timeline, eval- uation), briefly describe the participating sys- tems and summarise results for participating systems.

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.

Current captioning datasets focus on object-centric captions, describing the visible objects in the image, often ending up stating the obvious (for humans), e.g. “people eating food in a park”. Although these datasets are useful to evaluate the ability of Vision & Language models to recognize and describe visual content, they do not support controlled experiments involving model testing or fine-tuning, with more high-level captions, which humans find easy and natural to produce. For example, people often describe images based on the type of scene they depict (“people at a holiday resort”) and the actions they perform (“people having a picnic”). Such concepts are based on personal experience and contribute to forming common sense assumptions. We present the High-Level Dataset, a dataset extending 14997 images from the COCO dataset, aligned with a new set of 134,973 human-annotated (high-level) captions collected along three axes: scenes, actions and rationales. We further extend this dataset with confidence scores collected from an independent set of readers, as well as a set of narrative captions generated synthetically, by combining each of the three axes. We describe this dataset and analyse it extensively. We also present baseline results for the High-Level Captioning task.

Some applications of artificial intelligence make it desirable that logical formulae be converted computationally to comprehensible natural language sentences. As there are many logical equivalents to a given formula, finding the most suitable equivalent to be used as input for such a “logic-to-text” generation system is a difficult challenge. In this paper, we focus on the role of brevity: Are the shortest formulae the most suitable? We focus on propositional logic (PL), framing formula minimization (i.e., the problem of finding the shortest equivalent of a given formula) as a Quantified Boolean Formulae (QBFs) satisfiability problem. We experiment with several generators and selection strategies to prune the resulting candidates. We conduct exhaustive automatic and human evaluations of the comprehensibility and fluency of the generated texts. The results suggest that while, in many cases, minimization has a positive impact on the quality of the sentences generated, formula minimization may ultimately not be the best strategy.

This paper presents the Multitask, Multilingual, Multimodal Language Generation COST Action – Multi3Generation (CA18231), an interdisciplinary network of research groups working on different aspects of language generation. This “meta-paper” will serve as reference for citations of the Action in future publications. It presents the objectives, challenges and a the links for the achieved outcomes.

We propose VALSE (Vision And Language Structured Evaluation), a novel benchmark designed for testing general-purpose pretrained vision and language (V&L) models for their visio-linguistic grounding capabilities on specific linguistic phenomena. VALSE offers a suite of six tests covering various linguistic constructs. Solving these requires models to ground linguistic phenomena in the visual modality, allowing more fine-grained evaluations than hitherto possible. We build VALSE using methods that support the construction of valid foils, and report results from evaluating five widely-used V&L models. Our experiments suggest that current models have considerable difficulty addressing most phenomena. Hence, we expect VALSE to serve as an important benchmark to measure future progress of pretrained V&L models from a linguistic perspective, complementing the canonical task-centred V&L evaluations.

Image captioning models tend to describe images in an object-centric way, emphasising visible objects. But image descriptions can also abstract away from objects and describe the type of scene depicted. In this paper, we explore the potential of a state of the art Vision and Language model, VinVL, to caption images at the scene level using (1) a novel dataset which pairs images with both object-centric and scene descriptions. Through (2) an in-depth analysis of the effect of the fine-tuning, we show (3) that a small amount of curated data suffices to generate scene descriptions without losing the capability to identify object-level concepts in the scene; the model acquires a more holistic view of the image compared to when object-centric descriptions are generated. We discuss the parallels between these results and insights from computational and cognitive science research on scene perception.

Cookie banners are designed to request consent from website visitors for their personal data. Recent research suggest that a high percentage of cookie banners violate legal regulations as defined by the General Data Protection Regulation (GDPR) and the ePrivacy Directive. In this paper, we focus on language used in these cookie banners, and whether these violations can be automatically detected, or not. We make use of a small cookie banner dataset that is annotated by five experts for legal violations and test it with state of the art classification models, namely BERT, LEGAL-BERT, BART in a zero-shot setting and BERT with LIWC embeddings. Our results show that none of the models outperform the others in all classes, but in general, BERT and LEGAL-BERT provide the highest accuracy results (70%-97%). However, they are influenced by the small size and the unbalanced distributions in the dataset.

Current image description generation models do not transfer well to the task of describing human faces. To encourage the development of more human-focused descriptions, we developed a new data set of facial descriptions based on the CelebA image data set. We describe the properties of this data set, and present results from a face description generator trained on it, which explores the feasibility of using transfer learning from VGGFace/ResNet CNNs. Comparisons are drawn through both automated metrics and human evaluation by 76 English-speaking participants. The descriptions generated by the VGGFace-LSTM + Attention model are closest to the ground truth according to human evaluation whilst the ResNet-LSTM + Attention model obtained the highest CIDEr and CIDEr-D results (1.252 and 0.686 respectively). Together, the new data set and these experimental results provide data and baselines for future work in this area.

Multilingual language models such as mBERT have seen impressive cross-lingual transfer to a variety of languages, but many languages remain excluded from these models. In this paper, we analyse the effect of pre-training with monolingual data for a low-resource language that is not included in mBERT – Maltese – with a range of pre-training set ups. We conduct evaluations with the newly pre-trained models on three morphosyntactic tasks – dependency parsing, part-of-speech tagging, and named-entity recognition – and one semantic classification task – sentiment analysis. We also present a newly created corpus for Maltese, and determine the effect that the pre-training data size and domain have on the downstream performance. Our results show that using a mixture of pre-training domains is often superior to using Wikipedia text only. We also find that a fraction of this corpus is enough to make significant leaps in performance over Wikipedia-trained models. We pre-train and compare two models on the new corpus: a monolingual BERT model trained from scratch (BERTu), and a further pretrained multilingual BERT (mBERTu). The models achieve state-of-the-art performance on these tasks, despite the new corpus being considerably smaller than typically used corpora for high-resourced languages. On average, BERTu outperforms or performs competitively with mBERTu, and the largest gains are observed for higher-level tasks.

Logic-to-text generation is an important yet underrepresented area of natural language generation (NLG). In particular, most previous works on this topic lack sound evaluation. We address this limitation by building and evaluating a system that generates high-quality English text given a first-order logic (FOL) formula as input. We start by analyzing the performance of Ranta (2011)’s system. Based on this analysis, we develop an extended version of the system, which we name LoLa, that performs formula simplification based on logical equivalences and syntactic transformations. We carry out an extensive evaluation of LoLa using standard automatic metrics and human evaluation. We compare the results against a baseline and Ranta (2011)’s system. The results show that LoLa outperforms the other two systems in most aspects.

Noun-noun compounds (NNCs) occur frequently in the English language. Accurate NNC interpretation, i.e. determining the implicit relationship between the constituents of a NNC, is crucial for the advancement of many natural language processing tasks. Until now, computational NNC interpretation has been limited to approaches involving linguistic representations only. However, much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks. Our work is a novel comparison of linguistic and visuo-linguistic representations for the task of NNC interpretation. We frame NNC interpretation as a relation classification task, evaluating on a large, relationally-annotated NNC dataset. We combine distributional word vectors with image vectors to investigate how visual information can help improve NNC interpretation systems. We find that adding visual vectors increases classification performance on our dataset in many cases.

While powerful pre-trained language models have improved the fluency of text generation models, semantic adequacy -the ability to generate text that is semantically faithful to the input- remains an unsolved issue. In this paper, we introduce a novel automatic evaluation metric, Entity-Based Semantic Adequacy, which can be used to assess to what extent generation models that verbalise RDF (Resource Description Framework) graphs produce text that contains mentions of the entities occurring in the RDF input. This is important as RDF subject and object entities make up 2/3 of the input. We use our metric to compare 25 models from the WebNLG Shared Tasks and we examine correlation with results from human evaluations of semantic adequacy. We show that while our metric correlates with human evaluation scores, this correlation varies with the specifics of the human evaluation setup. This suggests that in order to measure the entity-based adequacy of generated texts, an automatic metric such as the one proposed here might be more reliable, as less subjective and more focused on correct verbalisation of the input, than human evaluation measures.

We investigate the reasoning ability of pretrained vision and language (V&L) models in two tasks that require multimodal integration: (1) discriminating a correct image-sentence pair from an incorrect one, and (2) counting entities in an image. We evaluate three pretrained V&L models on these tasks: ViLBERT, ViLBERT 12-in-1 and LXMERT, in zero-shot and finetuned settings. Our results show that models solve task (1) very well, as expected, since all models are pretrained on task (1). However, none of the pretrained V&L models is able to adequately solve task (2), our counting probe, and they cannot generalise to out-of-distribution quantities. We propose a number of explanations for these findings: LXMERT (and to some extent ViLBERT 12-in-1) show some evidence of catastrophic forgetting on task (1). Concerning our results on the counting probe, we find evidence that all models are impacted by dataset bias, and also fail to individuate entities in the visual input. While a selling point of pretrained V&L models is their ability to solve complex tasks, our findings suggest that understanding their reasoning and grounding capabilities requires more targeted investigations on specific phenomena.

Recent work has shown evidence that the knowledge acquired by multilingual BERT (mBERT) has two components: a language-specific and a language-neutral one. This paper analyses the relationship between them, in the context of fine-tuning on two tasks – POS tagging and natural language inference – which require the model to bring to bear different degrees of language-specific knowledge. Visualisations reveal that mBERT loses the ability to cluster representations by language after fine-tuning, a result that is supported by evidence from language identification experiments. However, further experiments on ‘unlearning’ language-specific representations using gradient reversal and iterative adversarial learning are shown not to add further improvement to the language-independent component over and above the effect of fine-tuning. The results presented here suggest that the process of fine-tuning causes a reorganisation of the model’s limited representational capacity, enhancing language-independent representations at the expense of language-specific ones.

This paper presents a novel scheme for the annotation of hate speech in corpora of Web 2.0 commentary. The proposed scheme is motivated by the critical analysis of posts made in reaction to news reports on the Mediterranean migration crisis and LGBTIQ+ matters in Malta, which was conducted under the auspices of the EU-funded C.O.N.T.A.C.T. project. Based on the realisation that hate speech is not a clear-cut category to begin with, appears to belong to a continuum of discriminatory discourse and is often realised through the use of indirect linguistic means, it is argued that annotation schemes for its detection should refrain from directly including the label ‘hate speech,’ as different annotators might have different thresholds as to what constitutes hate speech and what not. In view of this, we propose a multi-layer annotation scheme, which is pilot-tested against a binary ±hate speech classification and appears to yield higher inter-annotator agreement. Motivating the postulation of our scheme, we then present the MaNeCo corpus on which it will eventually be used; a substantial corpus of on-line newspaper comments spanning 10 years.

Maltese, the national language of Malta, is spoken by approximately 500,000 people. Speech processing for Maltese is still in its early stages of development. In this paper, we present the first spoken Maltese corpus designed purposely for Automatic Speech Recognition (ASR). The MASRI-HEADSET corpus was developed by the MASRI project at the University of Malta. It consists of 8 hours of speech paired with text, recorded by using short text snippets in a laboratory environment. The speakers were recruited from different geographical locations all over the Maltese islands, and were roughly evenly distributed by gender. This paper also presents some initial results achieved in baseline experiments for Maltese ASR using Sphinx and Kaldi. The MASRI HEADSET Corpus is publicly available for research/academic purposes.

End-to-end encoder-decoder approaches to data-to-text generation are often black boxes whose predictions are difficult to explain. Breaking up the end-to-end model into sub-modules is a natural way to address this problem. The traditional pre-neural Natural Language Generation (NLG) pipeline provides a framework for breaking up the end-to-end encoder-decoder. We survey recent papers that integrate traditional NLG submodules in neural approaches and analyse their explainability. Our survey is a first step towards building explainable neural NLG models.

The opaque nature of many machine learning techniques prevents the wide adoption of powerful information processing tools for high stakes scenarios. The emerging field eXplainable Artificial Intelligence (XAI) aims at providing justifications for automatic decision-making systems in order to ensure reliability and trustworthiness in the users. For achieving this vision, we emphasize the importance of a natural language textual modality as a key component for a future intelligent interactive agent. We outline the challenges of XAI and review a set of publications that work in this direction.

An ongoing debate in the NLG community concerns the best way to evaluate systems, with human evaluation often being considered the most reliable method, compared to corpus-based metrics. However, tasks involving subtle textual differences, such as style transfer, tend to be hard for humans to perform. In this paper, we propose an evaluation method for this task based on purposely-trained classifiers, showing that it better reflects system differences than traditional metrics such as BLEU.

Contextualized word embeddings have been replacing standard embeddings as the representational knowledge source of choice in NLP systems. Since a variety of biases have previously been found in standard word embeddings, it is crucial to assess biases encoded in their replacements as well. Focusing on BERT (Devlin et al., 2018), we measure gender bias by studying associations between gender-denoting target words and names of professions in English and German, comparing the findings with real-world workforce statistics. We mitigate bias by fine-tuning BERT on the GAP corpus (Webster et al., 2018), after applying Counterfactual Data Substitution (CDS) (Maudslay et al., 2019). We show that our method of measuring bias is appropriate for languages such as English, but not for languages with a rich morphology and gender-marking, such as German. Our results highlight the importance of investigating bias and mitigation techniques cross-linguistically,especially in view of the current emphasis on large-scale, multilingual language models.

Earlier research has shown that evaluation metrics based on textual similarity (e.g., BLEU, CIDEr, Meteor) do not correlate well with human evaluation scores for automatically generated text. We carried out an experiment with Chinese speakers, where we systematically manipulated image descriptions to contain different kinds of errors. Because our manipulated descriptions form minimal pairs with the reference descriptions, we are able to assess the impact of different kinds of errors on the perceived quality of the descriptions. Our results show that different kinds of errors elicit significantly different evaluation scores, even though all erroneous descriptions differ in only one character from the reference descriptions. Evaluation metrics based solely on textual similarity are unable to capture these differences, which (at least partially) explains their poor correlation with human judgments. Our work provides the foundations for future work, where we aim to understand why different errors are seen as more or less severe.

Natural Language Inference (NLI) is the task of determining the semantic relationship between a premise and a hypothesis. In this paper, we focus on the generation of hypotheses from premises in a multimodal setting, to generate a sentence (hypothesis) given an image and/or its description (premise) as the input. The main goals of this paper are (a) to investigate whether it is reasonable to frame NLI as a generation task; and (b) to consider the degree to which grounding textual premises in visual information is beneficial to generation. We compare different neural architectures, showing through automatic and human evaluation that entailments can indeed be generated successfully. We also show that multimodal models outperform unimodal models in this task, albeit marginally

Currently, there is little agreement as to how Natural Language Generation (NLG) systems should be evaluated. While there is some agreement regarding automatic metrics, there is a high degree of variation in the way that human evaluation is carried out. This paper provides an overview of how human evaluation is currently conducted, and presents a set of best practices, grounded in the literature. With this paper, we hope to contribute to the quality and consistency of human evaluations in NLG.

Inspired by Labov’s seminal work on stylisticvariation as a function of social stratification,we develop and compare neural models thatpredict a person’s presumed socio-economicstatus, obtained through distant supervision,from their writing style on social media. Thefocus of our work is on identifying the mostimportant stylistic parameters to predict socio-economic group. In particular, we show theeffectiveness of morpho-syntactic features aspredictors of style, in contrast to lexical fea-tures, which are good predictors of topic

Capturing semantic relations between sentences, such as entailment, is a long-standing challenge for computational semantics. Logic-based models analyse entailment in terms of possible worlds (interpretations, or situations) where a premise P entails a hypothesis H iff in all worlds where P is true, H is also true. Statistical models view this relationship probabilistically, addressing it in terms of whether a human would likely infer H from P. In this paper, we wish to bridge these two perspectives, by arguing for a visually-grounded version of the Textual Entailment task. Specifically, we ask whether models can perform better if, in addition to P and H, there is also an image (corresponding to the relevant “world” or “situation”). We use a multimodal version of the SNLI dataset (Bowman et al., 2015) and we compare “blind” and visually-augmented models of textual entailment. We show that visual information is beneficial, but we also conduct an in-depth error analysis that reveals that current multimodal models are not performing “grounding” in an optimal fashion.

We present a data resource which can be useful for research purposes on language grounding tasks in the context of geographical referring expression generation. The resource is composed of two data sets that encompass 25 different geographical descriptors and a set of associated graphical representations, drawn as polygons on a map by two groups of human subjects: teenage students and expert meteorologists.

In this paper we study empirically the validity of measures of referential success for referring expressions involving gradual properties. More specifically, we study the ability of several measures of referential success to predict the success of a user in choosing the right object, given a referring expression. Experimental results indicate that certain fuzzy measures of success are able to predict human accuracy in reference resolution. Such measures are therefore suitable for the estimation of the success or otherwise of a referring expression produced by a generation algorithm, especially in case the properties in a domain cannot be assumed to have crisp denotations.

Maltese is a morphologically rich language with a hybrid morphological system which features both concatenative and non-concatenative processes. This paper analyses the impact of this hybridity on the performance of machine learning techniques for morphological labelling and clustering. In particular, we analyse a dataset of morphologically related word clusters to evaluate the difference in results for concatenative and non-concatenative clusters. We also describe research carried out in morphological labelling, with a particular focus on the verb category. Two evaluations were carried out, one using an unseen dataset, and another one using a gold standard dataset which was manually labelled. The gold standard dataset was split into concatenative and non-concatenative to analyse the difference in results between the two morphological systems.

Image captioning has evolved into a core task for Natural Language Generation and has also proved to be an important testbed for deep learning approaches to handling multimodal representations. Most contemporary approaches rely on a combination of a convolutional network to handle image features, and a recurrent network to encode linguistic information. The latter is typically viewed as the primary “generation” component. Beyond this high-level characterisation, a CNN+RNN model supports a variety of architectural designs. The dominant model in the literature is one in which visual features encoded by a CNN are “injected” as part of the linguistic encoding process, driving the RNN’s linguistic choices. By contrast, it is possible to envisage an architecture in which visual and linguistic features are encoded separately, and merged at a subsequent stage. In this paper, we address two related questions: (1) Is direct injection the best way of combining multimodal information, or is a late merging alternative better for the image captioning task? (2) To what extent should a recurrent network be viewed as actually generating, rather than simply encoding, linguistic information?

System using BiLSTM and max pooling. Embedding is enhanced by POS, character and dependency info.

The automatic discovery and clustering of morphologically related words is an important problem with several practical applications. This paper describes the evaluation of word clusters carried out through crowd-sourcing techniques for the Maltese language. The hybrid (Semitic-Romance) nature of Maltese morphology, together with the fact that no large-scale lexical resources are available for Maltese, make this an interesting and challenging problem.

Starting in 2007, the field of natural language generation (NLG) has organised shared-task evaluation events every year, under the Generation Challenges umbrella. In the course of these shared tasks, a wealth of data has been created, along with associated task definitions and evaluation regimes. In other contexts too, sharable NLG data is now being created. In this paper, we describe the online repository that we have created as a one-stop resource for obtaining NLG task materials, both from Generation Challenges tasks and from other sources, where the set of materials provided for each task consists of (i) task definition, (ii) input and output data, (iii) evaluation software, (iv) documentation, and (v) publications reporting previous results.

This paper discusses the ongoing development of a new Maltese spell checker, highlighting the methodologies which would best suit such a language. We thus discuss several previous attempts, highlighting what we believe to be their weakest point: a lack of attention to context. Two developments are of particular interest, both of which concern the availability of language resources relevant to spellchecking: (i) the Maltese Language Resource Server (MLRS) which now includes a representative corpus of c. 100M words extracted from diverse documents including the Maltese Legislation, press releases and extracts from Maltese web-pages and (ii) an extensive and detailed corpus of spelling errors that was collected whilst part of the MLRS texts were being prepared. We describe the structure of these resources as well as the experimental approaches focused on context that we are now in a position to adopt. We describe the framework within which a variety of different approaches to spellchecking and evaluation will be carried out, and briefly discuss the first baseline system we have implemented. We conclude the paper with a roadmap for future improvements.

Notre société génère une masse d’information toujours croissante, que ce soit en médecine, en météorologie, etc. La méthode la plus employée pour analyser ces données est de les résumer sous forme graphique. Cependant, il a été démontré qu’un résumé textuel est aussi un mode de présentation efficace. L’objectif du prototype BT-45, développé dans le cadre du projet Babytalk, est de générer des résumés de 45 minutes de signaux physiologiques continus et d’événements temporels discrets en unité néonatale de soins intensifs (NICU). L’article présente l’aspect génération de texte de ce prototype. Une expérimentation clinique a montré que les résumés humains améliorent la prise de décision par rapport à l’approche graphique, tandis que les textes de BT-45 donnent des résultats similaires à l’approche graphique. Une analyse a identifié certaines des limitations de BT-45 mais en dépit de cellesci, notre travail montre qu’il est possible de produire automatiquement des résumés textuels efficaces de données complexes.