Language Resources Research Articles

Cyberbullying Text Identification based on Deep Learning and Transformer-based Language Models

In the contemporary digital age, social media platforms like Facebook, Twitter, and YouTube serve as vital channels for individuals to express ideas and connect with others. Despite fostering increased connectivity, these platforms have inadvertently given rise to negative behaviors, particularly cyberbullying. While extensive research has been conducted on high-resource languages such as English, there is a notable scarcity of resources for low-resource languages like Bengali, Arabic, Tamil, etc., particularly in terms of language modeling. This study addresses this gap by developing a cyberbullying text identification system called BullyFilterNeT tailored for social media texts, considering Bengali as a test case. The intelligent BullyFilterNeT system devised overcomes Out-of-Vocabulary (OOV) challenges associated with non-contextual embeddings and addresses the limitations of context-aware feature representations. To facilitate a comprehensive understanding, three non-contextual embedding models GloVe, FastText, and Word2Vec are developed for feature extraction in Bengali. These embedding models are utilized in the classification models, employing three statistical models (SVM, SGD, Libsvm), and four deep learning models (CNN, VDCNN, LSTM, GRU). Additionally, the study employs six transformer-based language models: mBERT, bELECTRA, IndicBERT, XML-RoBERTa, DistilBERT, and BanglaBERT, respectively to overcome the limitations of earlier models. Remarkably, BanglaBERT-based BullyFilterNeT achieves the highest accuracy of 88.04% in our test set, underscoring its effectiveness in cyberbullying text identification in the Bengali language.

Intertextuality of Brazilian comics: new meanings in literary adaptations

The dialogical principle of communicational objects draws our attention to intertextuality: the dialogue between texts. In comics, intertextuality is combined with the resources of a hybrid language: the verbal discourse, the technique or tool used, the imagery style and the page layout contribute to expanding the meanings of the adapted text. Given the possibilities of the language of comics, questions arise: which are the intertexts of the source works and how do adaptations use the language of comics to show them and include others, as every adaptation is a new work? For this investigation, we propose the analysis of two Brazilian comics: Auto da Barca do Inferno (2011) and Luzia (2021). In our theoretical-methodological journey, we work with notions of dialogism, intertextuality, and adaptation. We also consider the particular parameters of comics. From the analysis, we intend to understand how intertextuality is used in these adaptations.

A Privacy-Preserving Multilingual Comparable Corpus Construction Method in Internet of Things

With the expansion of the Internet of Things (IoT) and artificial intelligence (AI) technologies, multilingual scenarios are gradually increasing, and applications based on multilingual resources are also on the rise. In this process, apart from the need for the construction of multilingual resources, privacy protection issues like data privacy leakage are increasingly highlighted. Comparable corpus is important in multilingual language information processing in IoT. However, the multilingual comparable corpus concerning privacy preserving is rare, so there is an urgent need to construct a multilingual corpus resource. This paper proposes a method for constructing a privacy-preserving multilingual comparable corpus, taking Chinese–Uighur–Tibetan IoT based news as an example, and mapping the different language texts to a unified language vector space to avoid sensitive information, then calculates the similarity between different language texts and serves as a comparability index to construct comparable relations. Through the decision-making mechanism of minimizing the impossibility, it can identify a comparable corpus pair of multilingual texts based on chapter size to realize the construction of a privacy-preserving Chinese–Uighur–Tibetan comparable corpus (CUTCC). Evaluation experiments demonstrate the effectiveness of our proposed provable method, which outperforms in accuracy rate by 77%, recall rate by 34% and F value by 47.17%. The CUTCC provides valuable privacy-preserving data resources support and language service for multilingual situations in IoT.

A Minimal Computing Approach to Southern African Language Resources

A Minimal Computing Approach to Southern African Language Resources

Some bilingual couples speak lingua francas: personal pronoun indexicality in the light of positioning theory

ABSTRACT The paper focuses on the communication of three bilingual couples, each speaking a different lingua franca (LF). Positioning theory offers a methodological framework to explain language choice in interactive positioning within personal contacts. A comparative view of the couples’ storylines benefits from the differences between them. The analysis herein shows how the speakers are positioned in the subtle interplay of self-positioning and the positioning of the partner. The couples draw on their language resources and specific linguistic features which index language ideologies. Reasons for the choice and maintenance of a LF in the couple are informed by attitudinal factors which contribute to the discursive production of interpersonal positions. Findings outline multidimensional perspectives on LF couples by showing how different positions are created and how positioning is connected to the negotiation of a common language. This paper advances the debate on LF communication in intercultural relationships by demonstrating that positioning helps to investigate the ways in which diverse language resources intertwine to create different linguistic constellations and eventually improve our understanding of the performance in a third language.

MEDIALINGUISTICS AS A MODERN SCIENTIFIC DIRECTION

The features of media linguistics as a current scientific direction are considered. A distinction is made between the concepts of mass communication and the media. The identified features of the syntactic organization of the interview are determined by the desire to save language resources, which is predetermined, in particular, by the dialogical nature of the type of text under consideration. The term “media linguistics” itself contains two basic components in its structure – media (mass media) and linguistics. The object of study of the new discipline was a semiotically complicated text. The subject, in turn, is the study of the functioning of languages in the field of mass communication, and the text is characterized as a media text. Based on this, it becomes obvious that media linguistics is engaged in a comprehensive study of a certain area of speech use - the language of the media.

Machine learning based framework for fine-grained word segmentation and enhanced text normalization for low resourced language

In text applications, pre-processing is deemed as a significant parameter to enhance the outcomes of natural language processing (NLP) chores. Text normalization and tokenization are two pivotal procedures of text pre-processing that cannot be overstated. Text normalization refers to transforming raw text into scriptural standardized text, while word tokenization splits the text into tokens or words. Well defined normalization and tokenization approaches exist for most spoken languages in world. However, the world’s 10th most widely spoken language has been overlooked by the research community. This research presents improved text normalization and tokenization techniques for the Urdu language. For Urdu text normalization, multiple regular expressions and rules are proposed, including removing diuretics, normalizing single characters, separating digits, etc. While for word tokenization, core features are defined and extracted against each character of text. Machine learning model is considered with specified handcrafted rules to predict the space and to tokenize the text. This experiment is performed, while creating the largest human-annotated dataset composed in Urdu script covering five different domains. The results have been evaluated using precision, recall, F-measure, and accuracy. Further, the results are compared with state-of-the-art. The normalization approach produced 20% and tokenization approach achieved 6% improvement.

Improving neural machine translation for low resource languages through non-parallel corpora: a case study of Egyptian dialect to modern standard Arabic translation

Machine translation for low-resource languages poses significant challenges, primarily due to the limited availability of data. In recent years, unsupervised learning has emerged as a promising approach to overcome this issue by aiming to learn translations between languages without depending on parallel data. A wide range of methods have been proposed in the literature to address this complex problem. This paper presents an in-depth investigation of semi-supervised neural machine translation specifically focusing on translating Arabic dialects, particularly Egyptian, to Modern Standard Arabic. The study employs two distinct datasets: one parallel dataset containing aligned sentences in both dialects, and a monolingual dataset where the source dialect is not directly connected to the target language in the training data. Three different translation systems are explored in this study. The first is an attention-based sequence-to-sequence model that benefits from the shared vocabulary between the Egyptian dialect and Modern Arabic to learn word embeddings. The second is an unsupervised transformer model that depends solely on monolingual data, without any parallel data. The third system starts with the parallel dataset for an initial supervised learning phase and then incorporates the monolingual data during the training process.

TAXN: Translate Align Extract Normalize, a Multilingual Extraction Tool for Clinical Texts.

Several studies have shown that about 80% of the medical information in an electronic health record is only available through unstructured data. Resources such as medical terminologies in languages other than English are limited and restrain the NLP tools. We propose here to leverage English based resources in other languages using a combination of translation, word alignment, entity extraction and term normalization (TAXN). We implement this extraction pipeline in an open-source library called "medkit". We demonstrate the interest of this approach through a specific use-case: enriching a phenotypic dictionary for post-acute sequelae in COVID-19 (PASC). TAXN proved to be efficient to propose new synonyms of UMLS terms using a corpus of 70 articles in French with 356 terms enriched with at least one validated new synonym. This study was based on freely available deep-learning models.

Asturian cider in Madrid? Linguistic identity and the multilingual signage of Asturian restaurants in Madrid, Spain

ABSTRACT Migrant communities settle and appropriate spaces in their new home through deterritorialisation and reterritorialisation processes, which involve the reconceptualisation of the language and symbols of the homeland and those spoken and used in the diaspora. The public sphere is one of the most distinguishable places where this contextual relation is expressed. This paper examines how language and semiotic resources are used in the linguistic landscape of Asturian restaurants in the capital of Spain, Madrid, to represent spaces this community had in their motherland, Asturias (North-western Spain). Following a multimodal analysis of signage in ten Asturian restaurants, it analyses sign-genres, the distinction between symbolic and communicative functions of the signage, the language choice in authenticating places, and the placement of signs in the semiotic landscape.

Sonja Poulsen (2022). Collocations as a language resource: A functional and cognitive study in English phraseology

Sonja Poulsen (2022). Collocations as a language resource: A functional and cognitive study in English phraseology

Full-Fledged Access and Usability of Content in a Digital Cultural Heritage Library: Approaches, Paradigms, and Implementation

The current research presents an approach to the improvement of the librarian infrastructure via the integration of advanced technology and solutions aiming at an immediate, smooth, and continuous adaptation of the libraries in line with 21st-century needs. In particular, this article discusses the design and development of an innovative library environment ensuring descriptive cataloging, subject indexing, intelligent data searching, curation, and collection development aiming to meet the functional needs related to the increasingly growing users’ expectations. The effectiveness of the approach has been validated in the Regional Library “Peyo Yavorov”–Burgas, Bulgaria, an associated content provider in the Bulgarian National Interdisciplinary Research E-Infrastructure for Bulgarian Language and Cultural Heritage Resources and Technologies (CLaDA-BG). Testing of the system was made to assure its stability and reliability.

Hate Speech Detection in Limited Data Contexts Using Synthetic Data Generation

A growing body of work has focused on text classification methods for detecting the increasing amount of hate speech posted online. This progress has been limited to only a select number of highly resourced languages causing detection systems to either under-perform or not exist in limited data contexts. This is mostly caused by a lack of training data, which are expensive to collect and curate in these settings. In this work, we propose a data augmentation approach that addresses the problem of lack of data for online hate speech detection in limited data contexts using synthetic data generation techniques. Given a handful of hate speech examples in a high-resource language such as English, we present three methods to synthesize new examples of hate speech data in a target language that retains the hate sentiment in the original examples but transfers the hate targets. We apply our approach to generate training data for hate speech classification tasks in Hindi and Vietnamese. Our findings show that a model trained on synthetic data performs comparably to, and in some cases outperforms, a model trained only on the samples available in the target domain. This method can be adopted to bootstrap hate speech detection models from scratch in limited data contexts. As the growth of social media within these contexts continues to outstrip response efforts, this work furthers our capacities for detection, understanding, and response to hate speech. Disclaimer: This work contains terms that are offensive and hateful. These, however, cannot be avoided due to the nature of the work.

Moroccan Arabizi-to-Arabic conversion using rule-based transliteration and weighted Levenshtein algorithm

The rise of social media has contributed to the widespread of the Arabizi writing form, primarily used in colloquial communication. For Natural Language Processing (NLP) tools, processing texts in this form remains challenging due to the lack of suitable language resources. Additionally, there is a lack of standardized rules in the transliteration mapping between Arabizi and Arabic, resulting in variations across different dialectal groups. To address these limitations in the context of Moroccan Darija (MD), this work proposes a method for converting Arabizi to Arabic at the word level. This method involves a sequential combination of rule-based transliteration and weighted Levenshtein algorithm. The contributions of this approach include: (i) Building a large MD dataset that incorporates texts reflecting the characteristics of MD and the colloquial writing forms usually used in the Arabizi writing form. (ii) Generating transliteration rules tailored to MD. (iii) Adapting the edit costs within the weighted Levenshtein algorithm to enhance conversion performance. Successful tests have been conducted and the approach was applied to three datasets: two state-of-the-art Darija-Modern Standard Arabic (MSA) datasets and the MD dataset collected as part of this work. The proposed method achieved a Mean Reciprocal Rank (MRR) of 92.14% and an accuracy of 88.44%.

Are Emotions Conveyed Across Machine Translations? Establishing an Analytical Process for the Effectiveness of Multilingual Sentiment Analysis with Italian Text


 
 
 Abstract Natural language processing (NLP) is being widely used globally for a variety of value-creation tasks ranging from chat-bots and machine translations to sentiment and topic analysis and multilingual large language models (LLMs). However, most of the advances are initially implemented within the English language framework, and it takes time and resources to develop comparable resources in other languages. The advances in machine translations have enabled the rapid and effective conversion of content in global languages into English and vice-versa. This creates potential opportunities to apply English language NLP methods and tools to other languages via machine translations. However, although this idea is powerful, it needs to be validated and processes and best practices need to be developed and kept updated. The present research is an effort to contribute to the development of best practices and an evaluation framework. We present a systematic and repeatable state-of-the-art process to evaluate the viability of applying English language sentiment analysis tools to Italian text by using multiple English language machine translation mechanisms such that it can be easily extended to other languages.
 
 

Pristopi k izdelavi leksikalnih podatkovnih zbirk

The paper presents WordNet and EuroWordNet, lexical databases in which concepts are linked with lexical and semantic relations. In the past decade, similar lexical databases for other languages have been based on them, for which two approaches have mostly been used: the creation of independent wordnets and their subsequent integration into multilingual lexical databases, and the expansion of existing lexical databases into the target language by translating the entries and adopting relations between them. Due to the great automation potential of the network creation process and a uniform structure of the resulting databases, the latter approach has become very popular. This is why the given model was tested in this paper for its ability to create a Slovene wordnet, and the results were compared to its language-motivated counterpart acquired from the Fida corpus as well as other mono- and multilingual language resources for Slovene.

Linguistic disparities in cross-language automatic speech recognition transfer from Arabic to Tashlhiyt

Tashlhiyt is a low-resource language with respect to acoustic databases, language corpora, and speech technology tools, such as Automatic Speech Recognition (ASR) systems. This study investigates whether a method of cross-language re-use of ASR is viable for Tashlhiyt from an existing commercially-available system built for Arabic. The source and target language in this case have similar phonological inventories, but Tashlhiyt permits typologically rare phonological patterns, including vowelless words, while Arabic does not. We find systematic disparities in ASR transfer performance (measured as word error rate (WER) and Levenshtein distance) for Tashlhiyt across word forms and speaking style variation. Overall, performance was worse for casual speaking modes across the board. In clear speech, performance was lower for vowelless than for voweled words. These results highlight systematic speaking mode- and phonotactic-disparities in cross-language ASR transfer. They also indicate that linguistically-informed approaches to ASR re-use can provide more effective ways to adapt existing speech technology tools for low resource languages, especially when they contain typologically rare structures. The study also speaks to issues of linguistic disparities in ASR and speech technology more broadly. It can also contribute to understanding the extent to which machines are similar to, or different from, humans in mapping the acoustic signal to discrete linguistic representations.

“By the way I want to give you some masks”: exploring multimodal stance-taking in YouTube videos

Abstract While the effectiveness of facemasks against COVID-19 has now become largely uncontroversial, at the beginning of the global pandemic, wearers of facemasks were often the target of sometimes racially tinged attacks. Wearing facemasks (or not) became not just a question of science, but evolved into a more complex issue of social identity, morality and global citizenship embedded within the “tribal thinking” of mask-wearers and non-mask-wearers. This paper explores to what extent two bilingual YouTube influencers participated in either accentuating or softening of boundaries of the two “tribes” by embedding facemasks in their videos. Based on multimodal transcriptions of the two videos (Wang, Yilei, Dezheng Feng & Wing Y. J. Ho. 2021. Identity, lifestyle, and face-mask branding: A social semiotic multimodal discourse analysis. Multimodality & Society 1(2). 216–237), three moments were identified where facemasks were employed by the social actors to perform everyday activities, such as grocery shopping and family brunch. I then examine the interactional stances (Dubois, John W. 2007. The stance triangle. In Robert Englebretson (ed.), Stancetaking in discourse: Subjectivity, evaluation, interaction, 139–182. Amsterdam: John Benjamins) taken by the actors towards facemasks through language and other semiotic resources. By exploring their multimodal stance-taking, it is argued that the two YouTubers’ intercultural trajectories, their performances of authenticity, and their established influence on social media provided them unique means for participating in tribalizing discourses around facemasks by making perceived differences between different groups materials for cultural consumption. The paper concludes by discussing the opportunities and challenges of vernacular health communication through social media influencers.

The Element Selection and Analysis of Visual Transformation of Wenzhou Dialect

Wenzhou dialect, as a unique dialect, is not only known by the people all over China, but also plays a positive role in promoting the spread of Chinese culture and Wenzhou culture as Wenzhou merchant spread all over the country and even around the world. At the same time, like dialects in many areas of the country, Wenzhou dialect is also facing the possibility of shrinking the range of dissemination and audience, and shrinking or even disappearing of dialect user groups. In recent years, China’s language resources protection project is actively promoting the construction. While vigorously promoting the national standard language, it can also scientifically protect Chinese dialects and minority language resources, and promote the development and utilization of language resources. Dialect protection and dialect inheritance, as the most important resources, naturally belong to the core part, but the research on the visual transformation of dialects has rarely been involved. This paper tries to explore the selection and analysis of design elements in the visual transformation of Wenzhou dialect pronunciation.

A Study on Language Resources in COVID-19 Cartoons

During the COVID-19 lockdown, cartoonists played a crucial role in disseminating information and commentary through their art. This study conducted a thematic examination of 102 cartoons shared across Twitter, Facebook, and LinkedIn to understand societal reactions, critiques of authority’s handling of the crisis, and prominent issues of the pandemic. The analysis yielded three main themes: the impact and adaptation to the COVID-19 pandemic, the management of the crisis and COVID-19 awareness advocacy, and humorous portrayals of pandemic encounters. The study underscores the importance of cartoons as a valuable resource in future crisis communication, offering insights into effective and creative messaging strategies.