Skip to main content

Üveges István

Hungarian Academy of Sciences, Centre for Social Sciences,, Insitute of Political Science, Researcher

University of Szeged, Department of General Linguistics, Graduate Student

Followers

15

Following

41

Co-authors

5

Public Views

Researcher at the Centre for Social Sciences in the National Laboratory for Artificial Intelligence's (TK-MILAB) project on Sentiment Analysis, Ph.D. student in linguistics at the University of Szeged and a researcher in computational linguistics at MONTANA Knowledge Management Ltd.
Supervisors: Vincze Veronika and Ring Orsolya

less

Research Institute for Linguistics, Hungarian Academy of Sciences

György Rákosi

University of Debrecen

Zijian Győző Yang

ISTVAN DR FEKETE

Amity University

István Csertő

Hungarian Academy of Sciences

Hungarian Academy of Sciences, Centre for Social Sciences,

Mészáros Evelin

InterestsView All (7)

Uploads

Papers by Üveges István

A kontextusablakon kihajolni nem veszélyes: jogi szövegek hatékony szemantikus keresése

XXI. Magyar Számítógépes Nyelvészeti Konferencia, 2025

Tanulmányunkban egy szemantikus keresést támogató rendszert mutatunk be magyar bírósági határozat... more Tanulmányunkban egy szemantikus keresést támogató rendszert mutatunk be magyar bírósági határozatokon, amely képes az azonos
tényállású ügyek hatékony felismerésére és visszakeresésére. A kiértékelési rendszerünk alappillérét az úgynevezett „tényállásvázlatok” képezik,
amelyek tömören foglalják össze egy-egy ügy kulcsfontosságú tényeit, ezáltal lehetőséget biztosítva a releváns jogi precedensek gyors és pontos
azonosítására.
Vizsgálatunk során tizenkét korszerű szövegbeágyazási modellt vetettünk
össze, amelyek közül a Cohere, a Beijing Academy of Artificial Intelligence és az OpenAI legújabb fejlesztései teljesítettek a legjobban, különösen
a hosszú, zajos jogi szövegek feldolgozásában. A rövidebb kontextusablakkal bíró modellek esetében megvizsgáltunk több szegmentálási eljárást
is, köztük az egyszerű feldarabolást (chunking), az átlapolt feldarabolást
(striding) és a rövidebb utolsó chunk nagyobb hatását csökkentő (last
chunk scaling) technikákat, amelyek javították a hosszú szövegek megbízható kezelhetőségét.
A metodika egyaránt alkalmazható más Q&A jellegű rendszereknél is,
mivel pedig a legtöbb általunk kipróbált modell többnyelvű, így felhasználhatóság nem korlátozódik a magyarra.

From Fact Drafts to Operational Systems: Semantic Search in Legal Decisions Using Fact Drafts

Big Data and Cognitive Computing, 2024

This research paper presents findings from an investigation in the semantic similarity search tas... more This research paper presents findings from an investigation in the semantic similarity search task within the legal domain, using a corpus of 1172 Hungarian court decisions. The study establishes the groundwork for an operational semantic similarity search system designed to identify cases with comparable facts using preliminary legal fact drafts. Evaluating such systems often poses significant challenges, given the need for thorough document checks, which can be costly and limit evaluation reusability. To address this, the study employs manually created fact drafts for legal cases, enabling reliable ranking of original cases within retrieved documents and quantitative comparison of various vectorization methods. The study compares twelve different text embedding solutions (the most recent became available just a few weeks before the manuscript was written) identifying Cohere’s embed-multilingual-v3.0, Beijing Academy of Artificial Intelligence’s bge-m3, Jina AI’s jina-embeddings-v3, OpenAI’s text-embedding-3-large, and Microsoft’s multilingual-e5-large models as top performers. To overcome the transformer-based models’ context window limitation, we investigated chunking, striding, and last chunk scaling techniques, with last chunk scaling significantly improving embedding quality. The results suggest that the effectiveness of striding varies based on token count. Notably, employing striding with 16 tokens yielded optimal results, representing 3.125% of the context window size for the best-performing models. Results also suggested that from the models having 8192 token long context window the bge-m3 model is superior compared to jina-embeddings-v3 and text-embedding-3-large models in capturing the relevant parts of a document if the text contains significant amount of noise. The validity of the approach was evaluated and confirmed by legal experts. These insights led to an operational semantic search system for a prominent legal content provider.

Approaches to sentiment analysis of Hungarian political news at the sentence level

Language resources and evaluation, Mar 23, 2024

Automated sentiment analysis of textual data is one of the central and most challenging tasks in ... more Automated sentiment analysis of textual data is one of the central and most challenging tasks in political communication studies. However, the toolkits available are primarily for English texts and require contextual adaptation to produce valid results-especially concerning morphologically rich languages such as Hungarian. This study introduces (1) a new sentiment and emotion annotation framework that uses inductive approaches to identify emotions in the corpus and aggregate these emotions into positive, negative, and mixed sentiment categories, (2) a manually annotated sentiment data set with 5700 political news sentences, (3) a new Hungarian sentiment dictionary for political text analysis created via word embeddings, whose performance was compared with other available sentiment dictionaries. (4) Because of the limitations of sentiment analysis using dictionaries we have also applied various machine learning algorithms to analyze our dataset, (5) Last but not least to move towards state-of-the-art approaches, we have fine-tuned the Hungarian BERT-base model for sentiment analysis. Meanwhile, we have also tested how different pre-processing steps could affect the performance of machine-learning algorithms in the case of Hungarian texts.

Szövegaugmentálási módszerek összehasonlítása politikai szövegek szentimentanalízise során

XVIII. Magyar Számítógépes Nyelvészeti Konferencia, 2022

Cikkünkben bemutatjuk a gépi tanítási feladatokban gyakran előforduló kiegyensúlyozatlan tanítóha... more Cikkünkben bemutatjuk a gépi tanítási feladatokban gyakran előforduló kiegyensúlyozatlan tanítóhalmaz probléma egy lehetséges megoldását az alacsony elemszámú kategóriák szöveg-augmentálásával. Az összevethetőség érdekében egyszerű szövegaugmentálási technikákkal (EDA) és egy szóvektor alapú módszerrel is kísérletet teszünk. A módszerek hatékonyságát politikai doménbe tartozó szövegek szentimentelemzési feladatán teszteljük, amihez a TK-MILAB szentiment korpusz egy kisebb szeletét használjuk. Az alulreprezentált kategória bővítésével elért eredményeket a kiváltott F-érték változás függvényében értékeljük

Increasing access to legal information with unsupervised solutions

Hungarian Journal of Legal Studies, 2023

Access to justice is a significant area of legal research, especially for Socio-Legal studies. Th... more Access to justice is a significant area of legal research, especially for Socio-Legal studies. The main research topics of this area are economic or class differences, gender inequalities, or national and ethnic differences in access to justice. However, there is a less discussed aspect of access to justice: the differences between access to legal information and the differences in user groups in terms of comprehending and processing legal information. This is an important topic because there are significant differences among people’s abilities to process and understand legal texts, depending on whether we are dealing with a lawyer who is an expert in the given field, a non-expert lawyer, or a citizen with a low or zero (legal) educational level. The paper argues that unsupervised machine learning solutions can help even out these differences. It presents different unsupervised solutions, mainly clustering and topic modelling, which can help to increase access to legal information. Then we present a case study in which we examine these unsupervised tools in the processing of resolutions of the Central Bank in Hungary and anonymized court decisions. The paper argues that these tools can reveal the hidden contextual regularities in unstructured legal texts, facilitating the search for legal texts even for non-legal-experts.

Approaches to sentiment analysis of Hungarian political news at the sentence level

Language Resources and Evaluation, 2024

Automated sentiment analysis of textual data is one of the central and most challenging tasks in ... more Automated sentiment analysis of textual data is one of the central and most challenging tasks in political communication studies. However, the toolkits available are primarily for English texts and require contextual adaptation to produce valid results—especially concerning morphologically rich languages such as Hungarian. This study introduces (1) a new sentiment and emotion annotation framework that uses inductive approaches to identify emotions in the corpus and aggregate these emotions into positive, negative, and mixed sentiment categories, (2) a manually annotated sentiment data set with 5700 political news sentences, (3) a new Hungarian sentiment dictionary for political text analysis created via word embeddings, whose performance was compared with other available sentiment dictionaries. (4) Because of the limitations of sentiment analysis using dictionaries we have also applied various machine learning algorithms to analyze our dataset, (5) Last but not least to move towards state-of-the-art approaches, we have fine-tuned the Hungarian BERT-base model for sentiment analysis. Meanwhile, we have also tested how different pre-processing steps could affect the performance of machine-learning algorithms in the case of Hungarian texts.

Magyar Nyelvőr, 2024

One crucial aspect of access to justice and access to legal information is the comprehensibility ... more One crucial aspect of access to justice and access to legal information is the comprehensibility of legal text. The complexity and specialized terminology of legal language often prevents citizens from understanding legal texts and representing themselves effectively in legal proceedings. The new developments of Machine Learning, such as solutions based on Large Language Models, could represent a significant advancement in access to legal information, as they can transform complex legal texts into more straightforward, more understandable forms for laypeople. This paper attempts to exploit the capabilities of OpenAI’s GPT-4 model to produce automatic Plain Language transcriptions of legal texts. The experiment concerns four specific linguistic features, and the results are analyzed manually from both a legal and a linguistic point of view.

Can Triplet Loss Be Used for Multi-Label Few-Shot Classification? A Case Study

Information, Sep 22, 2023

This article is an open access article distributed under the terms and conditions of the Creative... more

Saving labeling cost by embracing Active Learning: a case study

Harnessing the power of Deep Learning is becoming commonplace nowadays, and the legal field is no... more Harnessing the power of Deep Learning is becoming commonplace nowadays, and the legal field is no exception. Most solutions embrace supervised approaches that require a lot of labeled data. Active Learning is a technique that exploits the already labeled data hence optimizing the time spent for labeling, saving time and costs. In this paper, the efficiency of Active Learning is investigated on a real-world hierarchical, multi-class multi-labeled dataset consisting of Hungarian legal documents from the taxation domain. Three different vectorization techniques (TF-IDF, fastText, BERT) are compared. The results suggest that by applying Active Learning, 40-60% less data is sufficient compared to random sampling and that lowering the batch size in an active learning cycle has a modest effect on reaching optimal performance.

Bírósági határozatok automatikus mondatszegmentálásának hatékonyságmérése

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

IEEE Access

The growing number of digitally accessible text corpora and the accelerating development of NLP t... more 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.

A magyar beszélt és írott nyelv különböző korpuszainak morfológiai és szófaji vizsgálata

Automatizálható a közérthető megfogalmazás? : Jog, számítógépes nyelvészet és pszicholingvisztika találkozása

StaffTalk: magyar nyelvű spontán beszélgetések korpusza

Szegedi Tudományegyetem, Informatikai Intézet, Jan 28, 2021

Szó, beszéd – avagy hogyan kommunikálunk egymásról

Magyar nyelvű spontán beszéd szemantikai-pragmatikai sajátságainak elemzése nagy méretű korpusz (StaffTalk) alapján

A dolgozatban bizonyos pragmatikai és szemantikai sajátságokat vizsgálunk magyar nyelvű, nagy mér... more A dolgozatban bizonyos pragmatikai és szemantikai sajátságokat vizsgálunk magyar nyelvű, nagy méretű spontánbeszéd-korpusz (StaffTalk) alapján. A vizsgálati korpusz is hétköznapi szituációkban, külső hatásoknak is kitett munkahelyi környezetben, spontán módon létrejött nyelvi tartalmakból áll, vagyis a kutatásban résztvevők szabadon megválaszthatták beszélgetésük tárgyát, hosszát és partnereit. A korpusz létrehozása során a hanganyagokat legépelték, majd azt követően számos szempont alapján annotálták. A jelen vizsgálatokat ezeket az annotációkat felhasználva végezzük el.

Üveges István: A mesterséges intelligencia közösségi médiában történő alkalmazásának társadalmi és politikai következményei

A mesterséges intelligencia és egyéb felforgató technológiák hatásainak átfogó vizsgálata, 2023

A mesterséges intelligenciát többen napjaink egyik leginkább felforgató technológiájaként határoz... more A mesterséges intelligenciát többen napjaink egyik leginkább felforgató technológiájaként határozzák meg. A tanulmány ezzel összefüggésben röviden ismerteti az adatalapú gazdasági modell kihatását a közösségi média felületekre, az adatbrókerek és más adatkezelők tevékenységének jellegét és kockázatait, valamint azokat a (már jelenleg is létező) manipulatív felhasználási módokat, ahogyan a mesterséges intelligencia alapú megoldások visszaélésekre adnak lehetőséget a magánélet védelme és a választások tisztasága kapcsán. Ennek célja, hogy ráirányítsa a figyelmet az ilyen formájú technológiai fejlődés jelentette kockázatokra, valamint feltárja azokat a folyamatokat, amelyek szabályozása a közeljövőben elengedhetetlenné válik

ParlaMint: Comparable Corpora of European Parliamentary Data

This paper outlines the ParlaMint project from the perspective of its goals, tasks, participants,... more This paper outlines the ParlaMint project from the perspective of its goals, tasks, participants, results and applications potential. The project produced language corpora from the sessions of the national parliaments of 17 countries, almost half a billion words in total. The corpora are split into COVID-related subcorpora (from November 2019) and reference corpora (to October 2019). The corpora are uniformly encoded according to the ParlaMint schema with the same Universal Dependencies linguistic annotations. Samples of the corpora and conversion scripts are available from the project's GitHub repository. The complete corpora are openly available via the CLARIN.SI repository 1 for download, and through the NoSketch Engine 2 and KonText 3 concordancers as well as through the Parlameter 4 interface for exploration and analysis.

Can Triplet Loss Be Used for Multi-Label Few-Shot Classification? A Case Study

by Üveges István, Renátó Vági, Andrea Megyeri, and Anna Fülöp

Few-shot learning is a deep learning subfield that is the focus of research nowadays. This paper ... more Few-shot learning is a deep learning subfield that is the focus of research nowadays. This paper addresses the research question of whether a triplet-trained Siamese network, initially designed for multi-class classification, can effectively handle multi-label classification. We conducted a case study to identify any limitations in its application. The experiments were conducted on a dataset containing Hungarian legal decisions of administrative agencies in tax matters belonging to a major legal content provider. We also tested how different Siamese embeddings compare on classifying a previously non-existing label on a binary and a multi-label setting. We found that triplet-trained Siamese networks can be applied to perform classification but with a sampling restriction during training. We also found that the overlap between labels affects the results negatively. The few-shot model, seeing only ten examples for each label, provided competitive results compared to models trained on tens of thousands of court decisions using tf-idf vectorization and logistic regression.

Named Entity Recognition in the Miskolc Legal Corpus

In this paper, a brief study will be presented with regard to the issue of Named Entity Recogniti... more In this paper, a brief study will be presented with regard to the issue of Named Entity Recognition (NER) in legal texts. To get an overall picture, we examined closely the output of two existing analysers: the "magyarlanc" linguistic processing toolkit [1] and a Named Entity Recognition system developed by the Natural Language Processing Group at the University of Szeged [2]. Firstly, short references are made to named entity recognition projects in the literature considered important in the current framework. Secondly, quantitative analyses of the data will be presented. At the end of the study, some problematic cases and potential solutions will be discussed which will be followed by the discussion of the future research.

A kontextusablakon kihajolni nem veszélyes: jogi szövegek hatékony szemantikus keresése

XXI. Magyar Számítógépes Nyelvészeti Konferencia, 2025

Tanulmányunkban egy szemantikus keresést támogató rendszert mutatunk be magyar bírósági határozat... more Tanulmányunkban egy szemantikus keresést támogató rendszert mutatunk be magyar bírósági határozatokon, amely képes az azonos
tényállású ügyek hatékony felismerésére és visszakeresésére. A kiértékelési rendszerünk alappillérét az úgynevezett „tényállásvázlatok” képezik,
amelyek tömören foglalják össze egy-egy ügy kulcsfontosságú tényeit, ezáltal lehetőséget biztosítva a releváns jogi precedensek gyors és pontos
azonosítására.
Vizsgálatunk során tizenkét korszerű szövegbeágyazási modellt vetettünk
össze, amelyek közül a Cohere, a Beijing Academy of Artificial Intelligence és az OpenAI legújabb fejlesztései teljesítettek a legjobban, különösen
a hosszú, zajos jogi szövegek feldolgozásában. A rövidebb kontextusablakkal bíró modellek esetében megvizsgáltunk több szegmentálási eljárást
is, köztük az egyszerű feldarabolást (chunking), az átlapolt feldarabolást
(striding) és a rövidebb utolsó chunk nagyobb hatását csökkentő (last
chunk scaling) technikákat, amelyek javították a hosszú szövegek megbízható kezelhetőségét.
A metodika egyaránt alkalmazható más Q&A jellegű rendszereknél is,
mivel pedig a legtöbb általunk kipróbált modell többnyelvű, így felhasználhatóság nem korlátozódik a magyarra.

From Fact Drafts to Operational Systems: Semantic Search in Legal Decisions Using Fact Drafts

Big Data and Cognitive Computing, 2024

This research paper presents findings from an investigation in the semantic similarity search tas... more This research paper presents findings from an investigation in the semantic similarity search task within the legal domain, using a corpus of 1172 Hungarian court decisions. The study establishes the groundwork for an operational semantic similarity search system designed to identify cases with comparable facts using preliminary legal fact drafts. Evaluating such systems often poses significant challenges, given the need for thorough document checks, which can be costly and limit evaluation reusability. To address this, the study employs manually created fact drafts for legal cases, enabling reliable ranking of original cases within retrieved documents and quantitative comparison of various vectorization methods. The study compares twelve different text embedding solutions (the most recent became available just a few weeks before the manuscript was written) identifying Cohere’s embed-multilingual-v3.0, Beijing Academy of Artificial Intelligence’s bge-m3, Jina AI’s jina-embeddings-v3, OpenAI’s text-embedding-3-large, and Microsoft’s multilingual-e5-large models as top performers. To overcome the transformer-based models’ context window limitation, we investigated chunking, striding, and last chunk scaling techniques, with last chunk scaling significantly improving embedding quality. The results suggest that the effectiveness of striding varies based on token count. Notably, employing striding with 16 tokens yielded optimal results, representing 3.125% of the context window size for the best-performing models. Results also suggested that from the models having 8192 token long context window the bge-m3 model is superior compared to jina-embeddings-v3 and text-embedding-3-large models in capturing the relevant parts of a document if the text contains significant amount of noise. The validity of the approach was evaluated and confirmed by legal experts. These insights led to an operational semantic search system for a prominent legal content provider.

Approaches to sentiment analysis of Hungarian political news at the sentence level

Language resources and evaluation, Mar 23, 2024

Automated sentiment analysis of textual data is one of the central and most challenging tasks in ... more Automated sentiment analysis of textual data is one of the central and most challenging tasks in political communication studies. However, the toolkits available are primarily for English texts and require contextual adaptation to produce valid results-especially concerning morphologically rich languages such as Hungarian. This study introduces (1) a new sentiment and emotion annotation framework that uses inductive approaches to identify emotions in the corpus and aggregate these emotions into positive, negative, and mixed sentiment categories, (2) a manually annotated sentiment data set with 5700 political news sentences, (3) a new Hungarian sentiment dictionary for political text analysis created via word embeddings, whose performance was compared with other available sentiment dictionaries. (4) Because of the limitations of sentiment analysis using dictionaries we have also applied various machine learning algorithms to analyze our dataset, (5) Last but not least to move towards state-of-the-art approaches, we have fine-tuned the Hungarian BERT-base model for sentiment analysis. Meanwhile, we have also tested how different pre-processing steps could affect the performance of machine-learning algorithms in the case of Hungarian texts.

Szövegaugmentálási módszerek összehasonlítása politikai szövegek szentimentanalízise során

XVIII. Magyar Számítógépes Nyelvészeti Konferencia, 2022

Cikkünkben bemutatjuk a gépi tanítási feladatokban gyakran előforduló kiegyensúlyozatlan tanítóha... more Cikkünkben bemutatjuk a gépi tanítási feladatokban gyakran előforduló kiegyensúlyozatlan tanítóhalmaz probléma egy lehetséges megoldását az alacsony elemszámú kategóriák szöveg-augmentálásával. Az összevethetőség érdekében egyszerű szövegaugmentálási technikákkal (EDA) és egy szóvektor alapú módszerrel is kísérletet teszünk. A módszerek hatékonyságát politikai doménbe tartozó szövegek szentimentelemzési feladatán teszteljük, amihez a TK-MILAB szentiment korpusz egy kisebb szeletét használjuk. Az alulreprezentált kategória bővítésével elért eredményeket a kiváltott F-érték változás függvényében értékeljük

Increasing access to legal information with unsupervised solutions

Hungarian Journal of Legal Studies, 2023

Access to justice is a significant area of legal research, especially for Socio-Legal studies. Th... more Access to justice is a significant area of legal research, especially for Socio-Legal studies. The main research topics of this area are economic or class differences, gender inequalities, or national and ethnic differences in access to justice. However, there is a less discussed aspect of access to justice: the differences between access to legal information and the differences in user groups in terms of comprehending and processing legal information. This is an important topic because there are significant differences among people’s abilities to process and understand legal texts, depending on whether we are dealing with a lawyer who is an expert in the given field, a non-expert lawyer, or a citizen with a low or zero (legal) educational level. The paper argues that unsupervised machine learning solutions can help even out these differences. It presents different unsupervised solutions, mainly clustering and topic modelling, which can help to increase access to legal information. Then we present a case study in which we examine these unsupervised tools in the processing of resolutions of the Central Bank in Hungary and anonymized court decisions. The paper argues that these tools can reveal the hidden contextual regularities in unstructured legal texts, facilitating the search for legal texts even for non-legal-experts.

Approaches to sentiment analysis of Hungarian political news at the sentence level

Language Resources and Evaluation, 2024

Automated sentiment analysis of textual data is one of the central and most challenging tasks in ... more Automated sentiment analysis of textual data is one of the central and most challenging tasks in political communication studies. However, the toolkits available are primarily for English texts and require contextual adaptation to produce valid results—especially concerning morphologically rich languages such as Hungarian. This study introduces (1) a new sentiment and emotion annotation framework that uses inductive approaches to identify emotions in the corpus and aggregate these emotions into positive, negative, and mixed sentiment categories, (2) a manually annotated sentiment data set with 5700 political news sentences, (3) a new Hungarian sentiment dictionary for political text analysis created via word embeddings, whose performance was compared with other available sentiment dictionaries. (4) Because of the limitations of sentiment analysis using dictionaries we have also applied various machine learning algorithms to analyze our dataset, (5) Last but not least to move towards state-of-the-art approaches, we have fine-tuned the Hungarian BERT-base model for sentiment analysis. Meanwhile, we have also tested how different pre-processing steps could affect the performance of machine-learning algorithms in the case of Hungarian texts.

Magyar Nyelvőr, 2024

One crucial aspect of access to justice and access to legal information is the comprehensibility ... more One crucial aspect of access to justice and access to legal information is the comprehensibility of legal text. The complexity and specialized terminology of legal language often prevents citizens from understanding legal texts and representing themselves effectively in legal proceedings. The new developments of Machine Learning, such as solutions based on Large Language Models, could represent a significant advancement in access to legal information, as they can transform complex legal texts into more straightforward, more understandable forms for laypeople. This paper attempts to exploit the capabilities of OpenAI’s GPT-4 model to produce automatic Plain Language transcriptions of legal texts. The experiment concerns four specific linguistic features, and the results are analyzed manually from both a legal and a linguistic point of view.

Can Triplet Loss Be Used for Multi-Label Few-Shot Classification? A Case Study

Information, Sep 22, 2023

This article is an open access article distributed under the terms and conditions of the Creative... more

Saving labeling cost by embracing Active Learning: a case study

Harnessing the power of Deep Learning is becoming commonplace nowadays, and the legal field is no... more Harnessing the power of Deep Learning is becoming commonplace nowadays, and the legal field is no exception. Most solutions embrace supervised approaches that require a lot of labeled data. Active Learning is a technique that exploits the already labeled data hence optimizing the time spent for labeling, saving time and costs. In this paper, the efficiency of Active Learning is investigated on a real-world hierarchical, multi-class multi-labeled dataset consisting of Hungarian legal documents from the taxation domain. Three different vectorization techniques (TF-IDF, fastText, BERT) are compared. The results suggest that by applying Active Learning, 40-60% less data is sufficient compared to random sampling and that lowering the batch size in an active learning cycle has a modest effect on reaching optimal performance.

Bírósági határozatok automatikus mondatszegmentálásának hatékonyságmérése

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

IEEE Access

The growing number of digitally accessible text corpora and the accelerating development of NLP t... more 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.

A magyar beszélt és írott nyelv különböző korpuszainak morfológiai és szófaji vizsgálata

Automatizálható a közérthető megfogalmazás? : Jog, számítógépes nyelvészet és pszicholingvisztika találkozása

StaffTalk: magyar nyelvű spontán beszélgetések korpusza

Szegedi Tudományegyetem, Informatikai Intézet, Jan 28, 2021

Szó, beszéd – avagy hogyan kommunikálunk egymásról

Magyar nyelvű spontán beszéd szemantikai-pragmatikai sajátságainak elemzése nagy méretű korpusz (StaffTalk) alapján

A dolgozatban bizonyos pragmatikai és szemantikai sajátságokat vizsgálunk magyar nyelvű, nagy mér... more A dolgozatban bizonyos pragmatikai és szemantikai sajátságokat vizsgálunk magyar nyelvű, nagy méretű spontánbeszéd-korpusz (StaffTalk) alapján. A vizsgálati korpusz is hétköznapi szituációkban, külső hatásoknak is kitett munkahelyi környezetben, spontán módon létrejött nyelvi tartalmakból áll, vagyis a kutatásban résztvevők szabadon megválaszthatták beszélgetésük tárgyát, hosszát és partnereit. A korpusz létrehozása során a hanganyagokat legépelték, majd azt követően számos szempont alapján annotálták. A jelen vizsgálatokat ezeket az annotációkat felhasználva végezzük el.

Üveges István: A mesterséges intelligencia közösségi médiában történő alkalmazásának társadalmi és politikai következményei

A mesterséges intelligencia és egyéb felforgató technológiák hatásainak átfogó vizsgálata, 2023

A mesterséges intelligenciát többen napjaink egyik leginkább felforgató technológiájaként határoz... more A mesterséges intelligenciát többen napjaink egyik leginkább felforgató technológiájaként határozzák meg. A tanulmány ezzel összefüggésben röviden ismerteti az adatalapú gazdasági modell kihatását a közösségi média felületekre, az adatbrókerek és más adatkezelők tevékenységének jellegét és kockázatait, valamint azokat a (már jelenleg is létező) manipulatív felhasználási módokat, ahogyan a mesterséges intelligencia alapú megoldások visszaélésekre adnak lehetőséget a magánélet védelme és a választások tisztasága kapcsán. Ennek célja, hogy ráirányítsa a figyelmet az ilyen formájú technológiai fejlődés jelentette kockázatokra, valamint feltárja azokat a folyamatokat, amelyek szabályozása a közeljövőben elengedhetetlenné válik

ParlaMint: Comparable Corpora of European Parliamentary Data

This paper outlines the ParlaMint project from the perspective of its goals, tasks, participants,... more This paper outlines the ParlaMint project from the perspective of its goals, tasks, participants, results and applications potential. The project produced language corpora from the sessions of the national parliaments of 17 countries, almost half a billion words in total. The corpora are split into COVID-related subcorpora (from November 2019) and reference corpora (to October 2019). The corpora are uniformly encoded according to the ParlaMint schema with the same Universal Dependencies linguistic annotations. Samples of the corpora and conversion scripts are available from the project's GitHub repository. The complete corpora are openly available via the CLARIN.SI repository 1 for download, and through the NoSketch Engine 2 and KonText 3 concordancers as well as through the Parlameter 4 interface for exploration and analysis.

Can Triplet Loss Be Used for Multi-Label Few-Shot Classification? A Case Study

by Üveges István, Renátó Vági, Andrea Megyeri, and Anna Fülöp

Few-shot learning is a deep learning subfield that is the focus of research nowadays. This paper ... more Few-shot learning is a deep learning subfield that is the focus of research nowadays. This paper addresses the research question of whether a triplet-trained Siamese network, initially designed for multi-class classification, can effectively handle multi-label classification. We conducted a case study to identify any limitations in its application. The experiments were conducted on a dataset containing Hungarian legal decisions of administrative agencies in tax matters belonging to a major legal content provider. We also tested how different Siamese embeddings compare on classifying a previously non-existing label on a binary and a multi-label setting. We found that triplet-trained Siamese networks can be applied to perform classification but with a sampling restriction during training. We also found that the overlap between labels affects the results negatively. The few-shot model, seeing only ten examples for each label, provided competitive results compared to models trained on tens of thousands of court decisions using tf-idf vectorization and logistic regression.

Named Entity Recognition in the Miskolc Legal Corpus

In this paper, a brief study will be presented with regard to the issue of Named Entity Recogniti... more In this paper, a brief study will be presented with regard to the issue of Named Entity Recognition (NER) in legal texts. To get an overall picture, we examined closely the output of two existing analysers: the "magyarlanc" linguistic processing toolkit [1] and a Named Entity Recognition system developed by the Natural Language Processing Group at the University of Szeged [2]. Firstly, short references are made to named entity recognition projects in the literature considered important in the current framework. Secondly, quantitative analyses of the data will be presented. At the end of the study, some problematic cases and potential solutions will be discussed which will be followed by the discussion of the future research.

Social and political implications of the use of artificial intelligence in social media

Constitutional Discourse Blog, 2023

https://constitutionaldiscourse.com/istvan-uveges-social-and-political-implications-of-the-use-of... more

Bírósági határozatok automatikus mondatszegmentálásának hatékonyságmérése

by Üveges István and Lakatos Dorina

A természetesnyelv-feldolgozási feladatok gyakori építőeleme a mondatokra történő szegmentálás, a... more A természetesnyelv-feldolgozási feladatok gyakori építőeleme a mondatokra történő szegmentálás, amely azonban a jogi szövegek esetében hagyományosan problémás terület. Jelen cikkünkben bemutatjuk a bírósági határozatok szegmentálására „hangolt“ szabályalapú eszközünket, összemérve azt pontosságban és futásidőben más magyar nyelvre alkalmazható mondatszegmentálókkal az eddigiekben mondatszegmentálás mérésére használt Szeged Treebank és UD korpuszokon, valamint egy csak magyar nyelvű bírósági határozatokat tartalmazó korpuszon. A szegmentálónk a Szeged Treebank korpuszon összességében a Stanza szegmentálóhoz hasonló eredményt ért el, azonban a jogi alkorpuszon már a legjobb modell volt, amely nem látta korábban ezt az adatot. A bírósági határozatokon a legjobbnak bizonyult a megközelítésünk. Megvizsgáltuk a tördelés ismeretének hatását a szegmentálókra, valamint a szövegek doménjeinek hatását a szegmentálókra