HuSpaCy: an industrial-strength Hungarian natural language processing toolkit

Orosz, György; Szántó, Zsolt; Berkecz, Péter; Szabó, G.; Farkas, Richárd

doi:10.48550/arxiv.2201.01956

Cited by 3 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This collection of keywords was revised and compiled into a keyword-to-topic mapping dictionary. Here, we used textacy [14] and the latest HuSpaCy (Orosz et al ., 2022) to collect key terms for the text set. The main steps of the “SZTAKI” annotation approach are as follows:Remove special audio transcript notations (e.g.…”

Section: Methodsmentioning

confidence: 99%

Identification of social scientifically relevant topics in an interview repository: a natural language processing experiment

Gárdos,

Egyed-Gergely,

Horváth

et al. 2023

View full text Add to dashboard Cite

PurposeThe present study is about generating metadata to enhance thematic transparency and facilitate research on interview collections at the Research Documentation Centre, Centre for Social Sciences (TK KDK) in Budapest. It explores the use of artificial intelligence (AI) in producing, managing and processing social science data and its potential to generate useful metadata to describe the contents of such archives on a large scale.Design/methodology/approachThe authors combined manual and automated/semi-automated methods of metadata development and curation. The authors developed a suitable domain-oriented taxonomy to classify a large text corpus of semi-structured interviews. To this end, the authors adapted the European Language Social Science Thesaurus (ELSST) to produce a concise, hierarchical structure of topics relevant in social sciences. The authors identified and tested the most promising natural language processing (NLP) tools supporting the Hungarian language. The results of manual and machine coding will be presented in a user interface.FindingsThe study describes how an international social scientific taxonomy can be adapted to a specific local setting and tailored to be used by automated NLP tools. The authors show the potential and limitations of existing and new NLP methods for thematic assignment. The current possibilities of multi-label classification in social scientific metadata assignment are discussed, i.e. the problem of automated selection of relevant labels from a large pool.Originality/valueInterview materials have not yet been used for building manually annotated training datasets for automated indexing of scientifically relevant topics in a data repository. Comparing various automated-indexing methods, this study shows a possible implementation of a researcher tool supporting custom visualizations and the faceted search of interview collections.

show abstract

Section: Methodsmentioning

confidence: 99%

Identification of social scientifically relevant topics in an interview repository: a natural language processing experiment

Gárdos,

Egyed-Gergely,

Horváth

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Perplexity remains a common choice in practical applications. However, some research suggests that coherence is the most effective method for measuring topic quality [47], with increased usage of this metric in recent studies. Despite the guidance provided by the above model evaluation methods, issues such as mixed topics, illogical topics, and indistinguishable topics can still arise.…”

Section: B Optimal Topics Number Selectionmentioning

confidence: 99%

Advancements of Artificial Intelligence Techniques in the Realm About Library and Information Subject—A Case Survey of Latent Dirichlet Allocation Method

Pan,

2023

IEEE Access

View full text Add to dashboard Cite

To investigate the advancements of artificial intelligence techniques in the realm of library and information subject, we have chosen the Latent Dirichlet Allocation method as a case study to explore its current study status and implementations. Traditional theme mining analyses utilize methods such as word frequency statistics, co-occurrence analysis, community detection, and citation analysis to capture external quantitative features of words or documents. In contrast, the Latent Dirichlet Allocation theme modelling method employs a three-layer Bayesian structure of document-topic-word to describe the themes of documents and the semantic relationships among words, enabling a better exploration of latent semantic information in text. This method plays a pivotal role in fine-grained knowledge extraction and analysis. We systematically review more than a decade of relevant literature in the realm about library and information subject. Through content analysis, we construct an analytical architecture for the implementation of the Latent Dirichlet Allocation method. This architecture, viewed from the perspective of the implementation process of Latent Dirichlet Allocation, comprehensively summarizes the core stages and technical challenges, including text pre-processing, model construction (i.e., theme model selection and optimal theme number determination), and model solving. Additionally, we provide a comprehensive overview of the current study status of the Latent Dirichlet Allocation method across various implementation domains, such as theme exploration, knowledge organization, academic evaluation, sentiment analysis, and recommendation study. Our findings indicate that the Latent Dirichlet Allocation method has formed a mature analytical process in the realm of library and information subject, with ongoing growth in study interest.

show abstract

“…For this purpose, we used the transformer-based pipeline 2 developed for the HuSpaCy [93] natural language processing toolkit for Hungarian. 3 Then, the emotion labels that the sentence has been annotated with at the clause level are determined for each sentence.…”

Section: A Data Selection and Corpus Statisticsmentioning

confidence: 99%

HunEmBERT: A Fine-Tuned BERT-Model for Classifying Sentiment and Emotion in Political Communication

Üveges

Ring

2023

IEEE Access

View full text Add to dashboard Cite

The growing number of digitally accessible text corpora and the accelerating development of NLP tools and methods (particularly the emergence of powerful large-scale language models) have allowed their widespread use in various classification tasks, including the vast field of sentiment analysis. However, these models must often be fine-tuned to perform this task efficiently. Therefore, we aimed to create a transformer-based fine-tuned model for the emotion and sentiment analysis of Hungarian political texts.The training data for the model were the manually annotated parliamentary speech texts from 2014 to 2018, which have the advantage of being rich in various emotions. The compiled corpus can be freely used for research purposes. In our work, we describe in detail the process of fine-tuning the Hungarian BERT model for sentiment and emotion classification, the performance achieved, and the typical classification errors, mainly due to a lack of recognition of pragmatic and other language use features by the fine-tuned models.INDEX TERMS Fine-tuned BERT-model, huBERT, emotion analysis, sentiment analysis, political communication.

show abstract

HuSpaCy: an industrial-strength Hungarian natural language processing toolkit

Cited by 3 publications

References 9 publications

Identification of social scientifically relevant topics in an interview repository: a natural language processing experiment

Identification of social scientifically relevant topics in an interview repository: a natural language processing experiment

Advancements of Artificial Intelligence Techniques in the Realm About Library and Information Subject—A Case Survey of Latent Dirichlet Allocation Method

HunEmBERT: A Fine-Tuned BERT-Model for Classifying Sentiment and Emotion in Political Communication

Contact Info

Product

Resources

About