Sentiment analysis is an important tool for aggregating patient voices, in order to provide targeted improvements in healthcare services. A prerequisite for this is the availability of in-domain data annotated for sentiment. This article documents an effort to add sentiment annotations to free-text comments in patient surveys collected by the Norwegian Institute of Public Health (NIPH). However, annotation can be a time-consuming and resource-intensive process, particularly when it requires domain expertise. We therefore also evaluate a possible alternative to human annotation, using large language models (LLMs) as annotators. We perform an extensive evaluation of the approach for two openly available pretrained LLMs for Norwegian, experimenting with different configurations of prompts and in-context learning, comparing their performance to human annotators. We find that even for zero-shot runs, models perform well above the baseline for binary sentiment, but still cannot compete with human annotators on the full dataset.

We investigate annotator variation for the novel task of Entity-Level Sentiment Analysis (ELSA) which annotates the aggregated sentiment directed towards volitional entities in a text. More specifically, we analyze the annotations of a newly constructed Norwegian ELSA dataset and release additional data with each annotator’s labels for the 247 entities in the dataset’s test split. We also perform a number of experiments prompting ChatGPT for these sentiment labels regarding each entity in the text and compare the generated annotations with the human labels. Cohen’s Kappa for agreement between the best LLM-generated labels and curated gold was 0.425, which indicates that these labels would not have high quality. Our analyses further investigate the errors that ChatGPT outputs, and compare them with the variations that we find among the 5 trained annotators that all annotated the same test data.

There is a large and growing body of literature on datasets created to facilitate the study of socio-political events of conflict and unrest. However, the datasets, and the approaches taken to create them, vary a lot depending on the type of research they are intended to support. For example, while scholars from natural language processing (NLP) tend to focus on annotating specific spans of text indicating various components of an event, scholars from the disciplines of political science and conflict studies tend to focus on creating databases that code an abstract but structured representation of the event, less tied to a specific source text.The survey presented in this paper aims to map out the current landscape of available event datasets within the domain of social and political conflict and unrest – both from the NLP and political science communities – offering a unified view of the work done across different disciplines.

In contrast to large text corpora, knowledge graphs (KG) provide dense and structured representations of factual information. This makes them attractive for systems that supplement or ground the knowledge found in pre-trained language models with an external knowledge source. This has especially been the case for classification tasks, where recent work has focused on creating pipeline models that retrieve information from KGs like ConceptNet as additional context. Many of these models consist of multiple components, and although they differ in the number and nature of these parts, they all have in common that for some given text query, they attempt to identify and retrieve a relevant subgraph from the KG. Due to the noise and idiosyncrasies often found in KGs, it is not known how current methods compare to a scenario where the aligned subgraph is completely relevant to the query. In this work, we try to bridge this knowledge gap by reviewing current approaches to text-to-KG alignment and evaluating them on two datasets where manually created graphs are available, providing insights into the effectiveness of current methods. We release our code for reproducibility.

We investigate in this paper how distributions of occupations with respect to gender is reflected in pre-trained language models. Such distributions are not always aligned to normative ideals, nor do they necessarily reflect a descriptive assessment of reality. In this paper, we introduce an approach for measuring to what degree pre-trained language models are aligned to normative and descriptive occupational distributions. To this end, we use official demographic information about gender–occupation distributions provided by the national statistics agencies of France, Norway, United Kingdom, and the United States. We manually generate template-based sentences combining gendered pronouns and nouns with occupations, and subsequently probe a selection of ten language models covering the English, French, and Norwegian languages. The scoring system we introduce in this work is language independent, and can be used on any combination of template-based sentences, occupations, and languages. The approach could also be extended to other dimensions of national census data and other demographic variables.

We present NorBench: a streamlined suite of NLP tasks and probes for evaluating Norwegian language models (LMs) on standardized data splits and evaluation metrics. We also introduce a range of new Norwegian language models (both encoder and encoder-decoder based). Finally, we compare and analyze their performance, along with other existing LMs, across the different benchmark tests of NorBench.

While modern masked language models (LMs) are trained on ever larger corpora, we here explore the effects of down-scaling training to a modestly-sized but representative, well-balanced, and publicly available English text source – the British National Corpus. We show that pre-training on this carefully curated corpus can reach better performance than the original BERT model. We argue that this type of corpora has great potential as a language modeling benchmark. To showcase this potential, we present fair, reproducible and data-efficient comparative studies of LMs, in which we evaluate several training objectives and model architectures and replicate previous empirical results in a systematic way. We propose an optimized LM architecture called LTG-BERT.

This paper provides a systematic analysis and comparison of the performance of state-of-the-art models on the task of fine-grained Arabic dialect identification using the MADAR parallel corpus. We test approaches based on pre-trained transformer language models in addition to Naive Bayes models with a rich set of various features. Through a comprehensive data- and error analysis, we provide valuable insights into the strengths and weaknesses of both approaches. We discuss which dialects are more challenging to differentiate, and identify potential sources of errors. Our analysis reveals an important problem with identical sentences across dialect classes in the test set of the MADAR-26 corpus, which may confuse any classifier. We also show that none of the tested approaches captures the subtle distinctions between closely related dialects.

This paper explores the use of masked language modeling (MLM) for data augmentation (DA), targeting structured sentiment analysis (SSA) for Norwegian based on a dataset of annotated reviews. Considering the limited resources for Norwegian language and the complexity of the annotation task, the aim is to investigate whether this approach to data augmentation can help boost the performance. We report on experiments with substituting words both inside and outside of sentiment annotations, and we also present an error analysis, discussing some of the potential pitfalls of using MLM-based DA for SSA, and suggest directions for future work.

Negation constitutes a challenging phenomenon for many natural language processing tasks, such as sentiment analysis (SA). In this paper we investigate the relationship between negation and sentiment in the context of Norwegian professional reviews. The first part of this paper includes a corpus study which investigates how negation is tied to sentiment in this domain, based on existing annotations. In the second part, we introduce NoReC-NegSynt, a synthetically augmented test set for negation and sentiment, to allow for a more detailed analysis of the role of negation in current neural SA models. This diagnostic test set, containing both clausal and non-clausal negation, allows for analyzing and comparing models’ abilities to treat several different types of negation. We also present a case-study, applying several neural SA models to the diagnostic data.

We present the Norwegian Anaphora Resolution Corpus (NARC), the first publicly available corpus annotated with anaphoric relations between noun phrases for Norwegian. The paper describes the annotated data for 326 documents in Norwegian Bokmål, together with inter-annotator agreement and discussions of relevant statistics. We also present preliminary modelling results which are comparable to existing corpora for other languages, and discuss relevant problems in relation to both modelling and the annotations themselves.

This paper demonstrates how a graph-based semantic parser can be applied to the task of structured sentiment analysis, directly predicting sentiment graphs from text. We advance the state of the art on 4 out of 5 standard benchmark sets. We release the source code, models and predictions.

In this paper, we introduce the first SemEval shared task on Structured Sentiment Analysis, for which participants are required to predict all sentiment graphs in a text, where a single sentiment graph is composed of a sentiment holder, target, expression and polarity. This new shared task includes two subtracks (monolingual and cross-lingual) with seven datasets available in five languages, namely Norwegian, Catalan, Basque, Spanish and English. Participants submitted their predictions on a held-out test set and were evaluated on Sentiment Graph F1 . Overall, the task received over 200 submissions from 32 participating teams. We present the results of the 15 teams that provided system descriptions and our own expanded analysis of the test predictions.

We present a qualitative analysis of the (potentially erroneous) outputs of contextualized embedding-based methods for detecting diachronic semantic change. First, we introduce an ensemble method outperforming previously described contextualized approaches. This method is used as a basis for an in-depth analysis of the degrees of semantic change predicted for English words across 5 decades. Our findings show that contextualized methods can often predict high change scores for words which are not undergoing any real diachronic semantic shift in the lexicographic sense of the term (or at least the status of these shifts is questionable). Such challenging cases are discussed in detail with examples, and their linguistic categorization is proposed. Our conclusion is that pre-trained contextualized language models are prone to confound changes in lexicographic senses and changes in contextual variance, which naturally stem from their distributional nature, but is different from the types of issues observed in methods based on static embeddings. Additionally, they often merge together syntactic and semantic aspects of lexical entities. We propose a range of possible future solutions to these issues.

In this paper we explore how a demographic distribution of occupations, along gender dimensions, is reflected in pre-trained language models. We give a descriptive assessment of the distribution of occupations, and investigate to what extent these are reflected in four Norwegian and two multilingual models. To this end, we introduce a set of simple bias probes, and perform five different tasks combining gendered pronouns, first names, and a set of occupations from the Norwegian statistics bureau. We show that language specific models obtain more accurate results, and are much closer to the real-world distribution of clearly gendered occupations. However, we see that none of the models have correct representations of the occupations that are demographically balanced between genders. We also discuss the importance of the training data on which the models were trained on, and argue that template-based bias probes can sometimes be fragile, and a simple alteration in a template can change a model’s behavior.

This paper explores the task of identifying the overall sentiment expressed towards volitional entities (persons and organizations) in a document - what we refer to as Entity-Level Sentiment Analysis (ELSA). While identifying sentiment conveyed towards an entity is well researched for shorter texts like tweets, we find little to no research on this specific task for longer texts with multiple mentions and opinions towards the same entity. This lack of research would be understandable if ELSA can be derived from existing tasks and models. To assess this, we annotate a set of professional reviews for their overall sentiment towards each volitional entity in the text. We sample from data already annotated for document-level, sentence-level, and target-level sentiment in a multi-domain review corpus, and our results indicate that there is no single proxy task that provides this overall sentiment we seek for the entities at a satisfactory level of performance. We present a suite of experiments aiming to assess the contribution towards ELSA provided by document-, sentence-, and target-level sentiment analysis, and provide a discussion of their shortcomings. We show that sentiment in our dataset is expressed not only with an entity mention as target, but also towards targets with a sentiment-relevant relation to a volitional entity. In our data, these relations extend beyond anaphoric coreference resolution, and our findings call for further research of the topic. Finally, we also present a survey of previous relevant work.

Fine-grained sentiment analysis attempts to extract sentiment holders, targets and polar expressions and resolve the relationship between them, but progress has been hampered by the difficulty of annotation. Targeted sentiment analysis, on the other hand, is a more narrow task, focusing on extracting sentiment targets and classifying their polarity. In this paper, we explore whether incorporating holder and expression information can improve target extraction and classification and perform experiments on eight English datasets. We conclude that jointly predicting target and polarity BIO labels improves target extraction, and that augmenting the input text with gold expressions generally improves targeted polarity classification. This highlights the potential importance of annotating expressions for fine-grained sentiment datasets. At the same time, our results show that performance of current models for predicting polar expressions is poor, hampering the benefit of this information in practice.

We present the ongoing NorLM initiative to support the creation and use of very large contextualised language models for Norwegian (and in principle other Nordic languages), including a ready-to-use software environment, as well as an experience report for data preparation and training. This paper introduces the first large-scale monolingual language models for Norwegian, based on both the ELMo and BERT frameworks. In addition to detailing the training process, we present contrastive benchmark results on a suite of NLP tasks for Norwegian. For additional background and access to the data, models, and software, please see: http://norlm.nlpl.eu

This paper introduces NorecNeg – the first annotated dataset of negation for Norwegian. Negation cues and their in-sentence scopes have been annotated across more than 11K sentences spanning more than 400 documents for a subset of the Norwegian Review Corpus (NoReC). In addition to providing in-depth discussion of the annotation guidelines, we also present a first set of benchmark results based on a graph-parsing approach.

Multilingual pretrained language models are rapidly gaining popularity in NLP systems for non-English languages. Most of these models feature an important corpus sampling step in the process of accumulating training data in different languages, to ensure that the signal from better resourced languages does not drown out poorly resourced ones. In this study, we train multiple multilingual recurrent language models, based on the ELMo architecture, and analyse both the effect of varying corpus size ratios on downstream performance, as well as the performance difference between monolingual models for each language, and broader multilingual language models. As part of this effort, we also make these trained models available for public use.

In this work we explore the effect of incorporating demographic metadata in a text classifier trained on top of a pre-trained transformer language model. More specifically, we add information about the gender of critics and book authors when classifying the polarity of book reviews, and the polarity of the reviews when classifying the genders of authors and critics. We use an existing data set of Norwegian book reviews with ratings by professional critics, which has also been augmented with gender information, and train a document-level sentiment classifier on top of a recently released Norwegian BERT-model. We show that gender-informed models obtain substantially higher accuracy, and that polarity-informed models obtain higher accuracy when classifying the genders of book authors. For this particular data set, we take this result as a confirmation of the gender bias in the underlying label distribution, but in other settings we believe a similar approach can be used for mitigating bias in the model.

Structured sentiment analysis attempts to extract full opinion tuples from a text, but over time this task has been subdivided into smaller and smaller sub-tasks, e.g., target extraction or targeted polarity classification. We argue that this division has become counterproductive and propose a new unified framework to remedy the situation. We cast the structured sentiment problem as dependency graph parsing, where the nodes are spans of sentiment holders, targets and expressions, and the arcs are the relations between them. We perform experiments on five datasets in four languages (English, Norwegian, Basque, and Catalan) and show that this approach leads to strong improvements over state-of-the-art baselines. Our analysis shows that refining the sentiment graphs with syntactic dependency information further improves results.

This paper presents NorNE, a manually annotated corpus of named entities which extends the annotation of the existing Norwegian Dependency Treebank. Comprising both of the official standards of written Norwegian (Bokmål and Nynorsk), the corpus contains around 600,000 tokens and annotates a rich set of entity types including persons, organizations, locations, geo-political entities, products, and events, in addition to a class corresponding to nominals derived from names. We here present details on the annotation effort, guidelines, inter-annotator agreement and an experimental analysis of the corpus using a neural sequence labeling architecture.

We here introduce NoReC_fine, a dataset for fine-grained sentiment analysis in Norwegian, annotated with respect to polar expressions, targets and holders of opinion. The underlying texts are taken from a corpus of professionally authored reviews from multiple news-sources and across a wide variety of domains, including literature, games, music, products, movies and more. We here present a detailed description of this annotation effort. We provide an overview of the developed annotation guidelines, illustrated with examples and present an analysis of inter-annotator agreement. We also report the first experimental results on the dataset, intended as a preliminary benchmark for further experiments.

Gender bias in models and datasets is widely studied in NLP. The focus has usually been on analysing how females and males express themselves, or how females and males are described. However, a less studied aspect is the combination of these two perspectives, how female and male describe the same or opposite gender. In this paper, we present a new gender annotated sentiment dataset of critics reviewing the works of female and male authors. We investigate if this newly annotated dataset contains differences in how the works of male and female authors are critiqued, in particular in terms of positive and negative sentiment. We also explore the differences in how this is done by male and female critics. We show that there are differences in how critics assess the works of authors of the same or opposite gender. For example, male critics rate crime novels written by females, and romantic and sentimental works written by males, more negatively.

This paper extends the task of probing sentence representations for linguistic insight in a multilingual domain. In doing so, we make two contributions: first, we provide datasets for multilingual probing, derived from Wikipedia, in five languages, viz. English, French, German, Spanish and Russian. Second, we evaluate six sentence encoders for each language, each trained by mapping sentence representations to English sentence representations, using sentences in a parallel corpus. We discover that cross-lingually mapped representations are often better at retaining certain linguistic information than representations derived from English encoders trained on natural language inference (NLI) as a downstream task.

We extend the well-known word analogy task to a one-to-X formulation, including one-to-none cases, when no correct answer exists. The task is cast as a relation discovery problem and applied to historical armed conflicts datasets, attempting to predict new relations of type ‘location:armed-group’ based on data about past events. As the source of semantic information, we use diachronic word embedding models trained on English news texts. A simple technique to improve diachronic performance in such task is demonstrated, using a threshold based on a function of cosine distance to decrease the number of false positives; this approach is shown to be beneficial on two different corpora. Finally, we publish a ready-to-use test set for one-to-X analogy evaluation on historical armed conflicts data.

We measure the intensity of diachronic semantic shifts in adjectives in English, Norwegian and Russian across 5 decades. This is done in order to test the hypothesis that evaluative adjectives are more prone to temporal semantic change. To this end, 6 different methods of quantifying semantic change are used. Frequency-controlled experimental results show that, depending on the particular method, evaluative adjectives either do not differ from other types of adjectives in terms of semantic change or appear to actually be less prone to shifting (particularly, to ‘jitter’-type shifting). Thus, in spite of many well-known examples of semantically changing evaluative adjectives (like ‘terrific’ or ‘incredible’), it seems that such cases are not specific to this particular type of words.

Neural methods for sentiment analysis have led to quantitative improvements over previous approaches, but these advances are not always accompanied with a thorough analysis of the qualitative differences. Therefore, it is not clear what outstanding conceptual challenges for sentiment analysis remain. In this work, we attempt to discover what challenges still prove a problem for sentiment classifiers for English and to provide a challenging dataset. We collect the subset of sentences that an (oracle) ensemble of state-of-the-art sentiment classifiers misclassify and then annotate them for 18 linguistic and paralinguistic phenomena, such as negation, sarcasm, modality, etc. Finally, we provide a case study that demonstrates the usefulness of the dataset to probe the performance of a given sentiment classifier with respect to linguistic phenomena.

This paper documents the creation of a large-scale dataset of evaluative sentences – i.e. both subjective and objective sentences that are found to be sentiment-bearing – based on mixed-domain professional reviews from various news-sources. We present both the annotation scheme and first results for classification experiments. The effort represents a step toward creating a Norwegian dataset for fine-grained sentiment analysis.

This paper explores the use of multi-task learning (MTL) for incorporating external knowledge in neural models. Specifically, we show how MTL can enable a BiLSTM sentiment classifier to incorporate information from sentiment lexicons. Our MTL set-up is shown to improve model performance (compared to a single-task set-up) on both English and Norwegian sentence-level sentiment datasets. The paper also introduces a new sentiment lexicon for Norwegian.

Encoders that generate representations based on context have, in recent years, benefited from adaptations that allow for pre-training on large text corpora. Earlier work on evaluating fixed-length sentence representations has included the use of ‘probing’ tasks, that use diagnostic classifiers to attempt to quantify the extent to which these encoders capture specific linguistic phenomena. The principle of probing has also resulted in extended evaluations that include relatively newer word-level pre-trained encoders. We build on probing tasks established in the literature and comprehensively evaluate and analyse – from a typological perspective amongst others – multilingual variants of existing encoders on probing datasets constructed for 6 non-English languages. Specifically, we probe each layer of a multiple monolingual RNN-based ELMo models, the transformer-based BERT’s cased and uncased multilingual variants, and a variant of BERT that uses a cross-lingual modelling scheme (XLM).

Recent years have witnessed a surge of publications aimed at tracing temporal changes in lexical semantics using distributional methods, particularly prediction-based word embedding models. However, this vein of research lacks the cohesion, common terminology and shared practices of more established areas of natural language processing. In this paper, we survey the current state of academic research related to diachronic word embeddings and semantic shifts detection. We start with discussing the notion of semantic shifts, and then continue with an overview of the existing methods for tracing such time-related shifts with word embedding models. We propose several axes along which these methods can be compared, and outline the main challenges before this emerging subfield of NLP, as well as prospects and possible applications.

In this paper, we empirically evaluate the utility of transfer and multi-task learning on a challenging semantic classification task: semantic interpretation of noun–noun compounds. Through a comprehensive series of experiments and in-depth error analysis, we show that transfer learning via parameter initialization and multi-task learning via parameter sharing can help a neural classification model generalize over a highly skewed distribution of relations. Further, we demonstrate how dual annotation with two distinct sets of relations over the same set of compounds can be exploited to improve the overall accuracy of a neural classifier and its F1 scores on the less frequent, but more difficult relations.

For decades, most self-respecting linguistic engineering initiatives have designed and implemented custom representations for various layers of, for example, morphological, syntactic, and semantic analysis. Despite occasional efforts at harmonization or even standardization, our field today is blessed with a multitude of ways of encoding and exchanging linguistic annotations of these types, both at the levels of ‘abstract syntax’, naming choices, and of course file formats. To a large degree, it is possible to work within and across design plurality by conversion, and often there may be good reasons for divergent design reflecting differences in use. However, it is likely that some abstract commonalities across choices of representation are obscured by more superficial differences, and conversely there is no obvious procedure to tease apart what actually constitute contentful vs. mere technical divergences. In this study, we seek to conceptually align three representations for common types of morpho-syntactic analysis, pinpoint what in our view constitute contentful differences, and reflect on the underlying principles and specific requirements that led to individual choices. We expect that a more in-depth understanding of these choices across designs may led to increased harmonization, or at least to more informed design of future representations.

This paper presents an open-source toolkit for negation detection. It identifies negation cues and their corresponding scope in either raw or parsed text using maximum-margin classification. The system design draws on best practice from the existing literature on negation detection, aiming for a simple and portable system that still achieves competitive performance. Pre-trained models and experimental results are provided for English.

Recent studies have shown that word embedding models can be used to trace time-related (diachronic) semantic shifts in particular words. In this paper, we evaluate some of these approaches on the new task of predicting the dynamics of global armed conflicts on a year-to-year basis, using a dataset from the conflict research field as the gold standard and the Gigaword news corpus as the training data. The results show that much work still remains in extracting ‘cultural’ semantic shifts from diachronic word embedding models. At the same time, we present a new task complete with an evaluation set and introduce the ‘anchor words’ method which outperforms previous approaches on this set.

This paper deals with using word embedding models to trace the temporal dynamics of semantic relations between pairs of words. The set-up is similar to the well-known analogies task, but expanded with a time dimension. To this end, we apply incremental updating of the models with new training texts, including incremental vocabulary expansion, coupled with learned transformation matrices that let us map between members of the relation. The proposed approach is evaluated on the task of predicting insurgent armed groups based on geographical locations. The gold standard data for the time span 1994–2010 is extracted from the UCDP Armed Conflicts dataset. The results show that the method is feasible and outperforms the baselines, but also that important work still remains to be done.

In this paper we present the Corpus of REcommendation STrength (CREST), a collection of HTML-formatted clinical guidelines annotated with the location of recommendations. Recommendations are labelled with an author-provided indicator of their strength of importance. As data was drawn from many disparate authors, we define a unified scheme of importance labels, and provide a mapping for each guideline. We demonstrate the utility of the corpus and its annotations in some initial measurements investigating the type of language constructions associated with strong and weak recommendations, and experiments into promising features for recommendation classification, both with respect to strong and weak labels, and to all labels of the unified scheme. An error analysis indicates that, while there is a strong relationship between lexical choices and strength labels, there can be substantial variance in the choices made by different authors.

The Linguistic Annotation Framework (LAF) provides an abstract data model for specifying interchange representations to ensure interoperability among different annotation formats. This paper describes an ongoing effort to adapt the LAF data model as the interchange representation in complex workflows as used in the Language Analysis Portal (LAP), an on-line and large-scale processing service that is developed as part of the Norwegian branch of the Common Language Resources and Technology Infrastructure (CLARIN) initiative. Unlike several related on-line processing environments, which predominantly instantiate a distributed architecture of web services, LAP achives scalability to potentially very large data volumes through integration with the Norwegian national e-Infrastructure, and in particular job sumission to a capacity compute cluster. This setup leads to tighter integration requirements and also calls for efficient, low-overhead communication of (intermediate) processing results with workflows. We meet these demands by coupling the LAF data model with a lean, non-redundant JSON-based interchange format and integration of an agile and performant NoSQL database, allowing parallel access from cluster nodes, as the central repository of linguistic annotation.

In this paper we describe and evaluate different statistical models for the task of realization ranking, i.e. the problem of discriminating between competing surface realizations generated for a given input semantics. Three models are trained and tested; an n-gram language model, a discriminative maximum entropy model using structural features, and a combination of these two. Our realization component forms part of a larger, hybrid MT system.