Text preprocessing is not only an essential step to prepare the corpus for modeling but also a key area that directly affects the natural language processing (NLP) application results. For instance, precise tokenization increases the accuracy of part-of-speech (POS) tagging, and retaining multiword expressions improves reasoning and machine translation. The text corpus needs to be appropriately preprocessed before it is ready to serve as the input to computer models. The preprocessing requirements depend on both the nature of the corpus and the NLP application itself, that is, what researchers would like to achieve from analyzing the data. Conventional text preprocessing practices generally suffice, but there exist situations where the text preprocessing needs to be customized for better analysis results. Hence, we discuss the pros and cons of several common text preprocessing methods: removing formatting, tokenization, text normalization, handling punctuation, removing stopwords, stemming and lemmatization, n-gramming, and identifying multiword expressions. Then, we provide examples of text datasets which require special preprocessing and how previous researchers handled the challenge. We expect this article to be a starting guideline on how to select and fine-tune text preprocessing methods.

A new framework for improving event detection is proposed that employs joint information in news media content and social networks, such as Twitter, to leverage detailed coverage of news media and the timeliness of social media. Specifically, a short text clustering method is employed to detect events from tweets, then the language model representations of the detected events are expanded using another set of events obtained from news articles published simultaneously. The expanded representations of events are employed as a new initialization of the clustering method to run another iteration and consequently enhance the event detection results. The proposed framework is evaluated using two datasets: a tweet dataset with event labels and a news dataset containing news articles published during the same time interval as the tweets. Experimental results show that the proposed framework improves the event detection results in terms of F1 measure compared to the results obtained from tweets only.

Many research topics in natural language processing (NLP), such as explanation generation, dialog modeling, or machine translation, require evaluation that goes beyond standard metrics like accuracy or F1 score toward a more human-centered approach. Therefore, understanding how to design user studies becomes increasingly important. However, few comprehensive resources exist on planning, conducting, and evaluating user studies for NLP, making it hard to get started for researchers without prior experience in the field of human evaluation. In this paper, we summarize the most important aspects of user studies and their design and evaluation, providing direct links to NLP tasks and NLP-specific challenges where appropriate. We (i) outline general study design, ethical considerations, and factors to consider for crowdsourcing, (ii) discuss the particularities of user studies in NLP, and provide starting points to select questionnaires, experimental designs, and evaluation methods that are tailored to the specific NLP tasks. Additionally, we offer examples with accompanying statistical evaluation code, to bridge the gap between theoretical guidelines and practical applications.

We are delighted to present the Special Issue on NLP Approaches to Offensive Content Online published in the Journal of Natural Language Engineering issue 29.6. We are happy to have received a total of 26 submissions to the special issue evidencing the interest of the NLP community in this topic. Our guest editorial board comprised of international experts in the field has worked hard to review all submissions over multiple rounds of peer review. Ultimately, we accepted nine articles to appear in this special issue.

Humans are naturally endowed with the ability to write in a particular style. They can, for instance, rephrase a formal letter in an informal way, convey a literal message with the use of figures of speech or edit a novel by mimicking the style of some well-known authors. Automating this form of creativity constitutes the goal of style transfer. As a natural language generation task, style transfer aims at rewriting existing texts, and specifically, it creates paraphrases that exhibit some desired stylistic attributes. From a practical perspective, it envisions beneficial applications, like chatbots that modulate their communicative style to appear empathetic, or systems that automatically simplify technical articles for a non-expert audience.

Several style-aware paraphrasing methods have attempted to tackle style transfer. A handful of surveys give a methodological overview of the field, but they do not support researchers to focus on specific styles. With this paper, we aim at providing a comprehensive discussion of the styles that have received attention in the transfer task. We organize them in a hierarchy, highlighting the challenges for the definition of each of them and pointing out gaps in the current research landscape. The hierarchy comprises two main groups. One encompasses styles that people modulate arbitrarily, along the lines of registers and genres. The other group corresponds to unintentionally expressed styles, due to an author’s personal characteristics. Hence, our review shows how these groups relate to one another and where specific styles, including some that have not yet been explored, belong in the hierarchy. Moreover, we summarize the methods employed for different stylistic families, hinting researchers towards those that would be the most fitting for future research.

Recent years have seen a growing number of publications that analyse Natural Language Understanding (NLU) datasets for superficial cues, whether they undermine the complexity of the tasks underlying those datasets and how they impact those models that are optimised and evaluated on this data. This structured survey provides an overview of the evolving research area by categorising reported weaknesses in models and datasets and the methods proposed to reveal and alleviate those weaknesses for the English language. We summarise and discuss the findings and conclude with a set of recommendations for possible future research directions. We hope that it will be a useful resource for researchers who propose new datasets to assess the suitability and quality of their data to evaluate various phenomena of interest, as well as those who propose novel NLU approaches, to further understand the implications of their improvements with respect to their model’s acquired capabilities.

Recent research has reported that standard fine-tuning approaches can be unstable due to being prone to various sources of randomness, including but not limited to weight initialization, training data order, and hardware. Such brittleness can lead to different evaluation results, prediction confidences, and generalization inconsistency of the same models independently fine-tuned under the same experimental setup. Our paper explores this problem in natural language inference, a common task in benchmarking practices, and extends the ongoing research to the multilingual setting. We propose six novel textual entailment and broad-coverage diagnostic datasets for French, German, and Swedish. Our key findings are that the mBERT model demonstrates fine-tuning instability for categories that involve lexical semantics, logic, and predicate-argument structure and struggles to learn monotonicity, negation, numeracy, and symmetry. We also observe that using extra training data only in English can enhance the generalization performance and fine-tuning stability, which we attribute to the cross-lingual transfer capabilities. However, the ratio of particular features in the additional training data might rather hurt the performance for model instances. We are publicly releasing the datasets, hoping to foster the diagnostic investigation of language models (LMs) in a cross-lingual scenario, particularly in terms of benchmarking, which might promote a more holistic understanding of multilingualism in LMs and cross-lingual knowledge transfer.

We present novel methods for detecting lexical entailment in a fully rule-based and explainable fashion, by automatic construction of semantic graphs, in any language for which a crowd-sourced dictionary with sufficient coverage and a dependency parser of sufficient accuracy are available. We experiment and evaluate on both the Semeval-2020 lexical entailment task (Glavaš et al. (2020). Proceedings of the Fourteenth Workshop on Semantic Evaluation, pp. 24–35) and the SherLIiC lexical inference dataset of typed predicates (Schmitt and Schütze (2019). Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 902–914). Combined with top-performing systems, our method achieves improvements over the previous state-of-the-art on both benchmarks. As a standalone system, it offers a fully interpretable model of lexical entailment that makes detailed error analysis possible, uncovering future directions for improving both the semantic parsing method and the inference process on semantic graphs. We release all components of our system as open source software.

The OffensEval shared tasks organized as part of SemEval-2019–2020 were very popular, attracting over 1300 participating teams. The two editions of the shared task helped advance the state of the art in offensive language identification by providing the community with benchmark datasets in Arabic, Danish, English, Greek, and Turkish. The datasets were annotated using the OLID hierarchical taxonomy, which since then has become the de facto standard in general offensive language identification research and was widely used beyond OffensEval. We present a survey of OffensEval and related competitions, and we discuss the main lessons learned. We further evaluate the performance of Large Language Models (LLMs), which have recently revolutionalized the field of Natural Language Processing. We use zero-shot prompting with six popular LLMs and zero-shot learning with two task-specific fine-tuned BERT models, and we compare the results against those of the top-performing teams at the OffensEval competitions. Our results show that while some LMMs such as Flan-T5 achieve competitive performance, in general LLMs lag behind the best OffensEval systems.

We give an in-depth account of compositional matrix-space models (CMSMs), a type of generic models for natural language, wherein compositionality is realized via matrix multiplication. We argue for the structural plausibility of this model and show that it is able to cover and combine various common compositional natural language processing approaches. Then, we consider efficient task-specific learning methods for training CMSMs and evaluate their performance in compositionality prediction and sentiment analysis.

Property inference involves predicting properties for a word from its distributional representation. We focus on human-generated resources that link words to their properties and on the task of predicting these properties for unseen words. We introduce the use of label propagation, a semi-supervised machine learning approach, for this task and, in the first systematic study of models for this task, find that label propagation achieves state-of-the-art results. For more variety in the kinds of properties tested, we introduce two new property datasets.

Learning idiomatic expressions is seen as one of the most challenging stages in second-language learning because of their unpredictable meaning. A similar situation holds for their identification within natural language processing applications such as machine translation and parsing. The lack of high-quality usage samples exacerbates this challenge not only for humans but also for artificial intelligence systems. This article introduces a gamified crowdsourcing approach for collecting language learning materials for idiomatic expressions; a messaging bot is designed as an asynchronous multiplayer game for native speakers who compete with each other while providing idiomatic and nonidiomatic usage examples and rating other players’ entries. As opposed to classical crowd-processing annotation efforts in the field, for the first time in the literature, a crowd-creating & crowd-rating approach is implemented and tested for idiom corpora construction. The approach is language-independent and evaluated on two languages in comparison to traditional data preparation techniques in the field. The reaction of the crowd is monitored under different motivational means (namely, gamification affordances and monetary rewards). The results reveal that the proposed approach is powerful in collecting the targeted materials, and although being an explicit crowdsourcing approach, it is found entertaining and useful by the crowd. The approach has been shown to have the potential to speed up the construction of idiom corpora for different natural languages to be used as second-language learning material, training data for supervised idiom identification systems, or samples for lexicographic studies.

Hate speech is any kind of communication that attacks a person or a group based on their characteristics, such as gender, religion and race. Due to the availability of online platforms where people can express their (hateful) opinions, the amount of hate speech is steadily increasing that often leads to offline hate crimes. This paper focuses on understanding and detecting hate speech in underground hacking and extremist forums where cybercriminals and extremists, respectively, communicate with each other, and some of them are associated with criminal activity. Moreover, due to the lengthy posts, it would be beneficial to identify the specific span of text containing hateful content in order to assist site moderators with the removal of hate speech. This paper describes a hate speech dataset composed of posts extracted from HackForums, an online hacking forum, and Stormfront and Incels.co, two extremist forums. We combined our dataset with a Twitter hate speech dataset to train a multi-platform classifier. Our evaluation shows that a classifier trained on multiple sources of data does not always improve the performance compared to a mono-platform classifier. Finally, this is the first work on extracting hate speech spans from longer texts. The paper fine-tunes BERT (Bidirectional Encoder Representations from Transformers) and adopts two approaches – span prediction and sequence labelling. Both approaches successfully extract hateful spans and achieve an F1-score of at least 69%.

This study describes a Natural Language Processing (NLP) toolkit, as the first contribution of a larger project, for an under-resourced language—Urdu. In previous studies, standard NLP toolkits have been developed for English and many other languages. There is also a dire need for standard text processing tools and methods for Urdu, despite it being widely spoken in different parts of the world with a large amount of digital text being readily available. This study presents the first version of the UNLT (Urdu Natural Language Toolkit) which contains three key text processing tools required for an Urdu NLP pipeline; word tokenizer, sentence tokenizer, and part-of-speech (POS) tagger. The UNLT word tokenizer employs a morpheme matching algorithm coupled with a state-of-the-art stochastic n-gram language model with back-off and smoothing characteristics for the space omission problem. The space insertion problem for compound words is tackled using a dictionary look-up technique. The UNLT sentence tokenizer is a combination of various machine learning, rule-based, regular-expressions, and dictionary look-up techniques. Finally, the UNLT POS taggers are based on Hidden Markov Model and Maximum Entropy-based stochastic techniques. In addition, we have developed large gold standard training and testing data sets to improve and evaluate the performance of new techniques for Urdu word tokenization, sentence tokenization, and POS tagging. For comparison purposes, we have compared the proposed approaches with several methods. Our proposed UNLT, the training and testing data sets, and supporting resources are all free and publicly available for academic use.

In this article, we present the Construction Grammar Conceptual Network method, developed for identifying lexical similarity and word sense discrimination in a syntactically tagged corpus, based on the cognitive linguistic assumption that coordination construction instantiates conceptual relatedness. This graph analysis method projects a semantic value onto a given coordinated syntactic dependency and constructs a second-order lexical network of lexical collocates with a high co-occurrence measure. The subsequent process of clustering and pruning the graph reveals lexical communities with high conceptual similarity, which are interpreted as associated senses of the source lexeme. We demonstrate the theory and its application to the task of identifying the conceptual structure and different meanings of nouns, adjectives and verbs using examples from different corpora, and explain the modulating effects of linguistic and graph parameters. This graph approach is based on syntactic dependency processing and can be used as a complementary method to other contemporary natural language processing resources to enrich semantic tasks such as word disambiguation, domain relatedness, sense structure, identification of synonymy, metonymy, and metaphoricity, as well as to automate comprehensive meta-reasoning about languages and identify cross/intra-cultural discourse variations of prototypical conceptualization patterns and knowledge representations. As a contribution, we provide a web-based app at http://emocnet.uniri.hr/.

In multiple-choice exams, students select one answer from among typically four choices and can explain why they made that particular choice. Students are good at understanding natural language questions and based on their domain knowledge can easily infer the question’s answer by “connecting the dots” across various pertinent facts. Considering automated reasoning for elementary science question answering, we address the novel task of generating explanations for answers from human-authored facts. For this, we examine the practically scalable framework of feature-rich support vector machines leveraging domain-targeted, hand-crafted features. Explanations are created from a human-annotated set of nearly 5000 candidate facts in the WorldTree corpus. Our aim is to obtain better matches for valid facts of an explanation for the correct answer of a question over the available fact candidates. To this end, our features offer a comprehensive linguistic and semantic unification paradigm. The machine learning problem is the preference ordering of facts, for which we test pointwise regression versus pairwise learning-to-rank. Our contributions, originating from comprehensive evaluations against nine existing systems, are (1) a case study in which two preference ordering approaches are systematically compared, and where the pointwise approach is shown to outperform the pairwise approach, thus adding to the existing survey of observations on this topic; (2) since our system outperforms a highly-effective TF-IDF-based IR technique by 3.5 and 4.9 points on the development and test sets, respectively, it demonstrates some of the further task improvement possibilities (e.g., in terms of an efficient learning algorithm, semantic features) on this task; (3) it is a practically competent approach that can outperform some variants of BERT-based reranking models; and (4) the human-engineered features make it an interpretable machine learning model for the task.

We introduce a generic, language-independent method to collect a large percentage of offensive and hate tweets regardless of their topics or genres. We harness the extralinguistic information embedded in the emojis to collect a large number of offensive tweets. We apply the proposed method on Arabic tweets and compare it with English tweets—analyzing key cultural differences. We observed a constant usage of these emojis to represent offensiveness throughout different timespans on Twitter. We manually annotate and publicly release the largest Arabic dataset for offensive, fine-grained hate speech, vulgar, and violence content. Furthermore, we benchmark the dataset for detecting offensiveness and hate speech using different transformer architectures and perform in-depth linguistic analysis. We evaluate our models on external datasets—a Twitter dataset collected using a completely different method, and a multi-platform dataset containing comments from Twitter, YouTube, and Facebook, for assessing generalization capability. Competitive results on these datasets suggest that the data collected using our method capture universal characteristics of offensive language. Our findings also highlight the common words used in offensive communications, common targets for hate speech, specific patterns in violence tweets, and pinpoint common classification errors that can be attributed to limitations of NLP models. We observe that even state-of-the-art transformer models may fail to take into account culture, background, and context or understand nuances present in real-world data such as sarcasm.

Convolutional sequence to sequence (CNN seq2seq) models have met success in abstractive summarization. However, their outputs often contain repetitive word sequences and logical inconsistencies, limiting the practicality of their application. In this paper, we find the reasons behind the repetition problem in CNN-based abstractive summarization through observing the attention map between the summaries with repetition and their corresponding source documents and mitigate the repetition problem. We propose to reduce the repetition in summaries by attention filter mechanism (ATTF) and sentence-level backtracking decoder (SBD), which dynamically redistributes attention over the input sequence as the output sentences are generated. The ATTF can record previously attended locations in the source document directly and prevent the decoder from attending to these locations. The SBD prevents the decoder from generating similar sentences more than once via backtracking at test. The proposed model outperforms the baselines in terms of ROUGE score, repeatedness, and readability. The results show that this approach generates high-quality summaries with minimal repetition and makes the reading experience better.

Previouswork in slogan generation focused on utilising slogan skeletons mined from existing slogans. While some generated slogans can be catchy, they are often not coherent with the company’s focus or style across their marketing communications because the skeletons are mined from other companies’ slogans. We propose a sequence-to-sequence (seq2seq) Transformer model to generate slogans from a brief company description. A naïve seq2seq model fine-tuned for slogan generation is prone to introducing false information. We use company name delexicalisation and entity masking to alleviate this problem and improve the generated slogans’ quality and truthfulness. Furthermore, we apply conditional training based on the first words’ part-of-speech tag to generate syntactically diverse slogans. Our best model achieved a ROUGE-1/-2/-L $\mathrm{F}_1$ score of 35.58/18.47/33.32. Besides, automatic and human evaluations indicate that our method generates significantly more factual, diverse and catchy slogans than strong long short-term memory and Transformer seq2seq baselines.

Natural language processing (NLP) technologies and applications in legal text processing are gaining momentum. Being one of the most prominent tasks in NLP, named-entity recognition (NER) can substantiate a great convenience for NLP in law due to the variety of named entities in the legal domain and their accentuated importance in legal documents. However, domain-specific NER models in the legal domain are not well studied. We present a NER model for Turkish legal texts with a custom-made corpus as well as several NER architectures based on conditional random fields and bidirectional long-short-term memories (BiLSTMs) to address the task. We also study several combinations of different word embeddings consisting of GloVe, Morph2Vec, and neural network-based character feature extraction techniques either with BiLSTM or convolutional neural networks. We report 92.27% F1 score with a hybrid word representation of GloVe and Morph2Vec with character-level features extracted with BiLSTM. Being an agglutinative language, the morphological structure of Turkish is also considered. To the best of our knowledge, our work is the first legal domain-specific NER study in Turkish and also the first study for an agglutinative language in the legal domain. Thus, our work can also have implications beyond the Turkish language.