CLARIN-Emo: Training Emotion Recognition Models Using Human Annotation and ChatGPT

Ngô, Anh Dung; Kocoń, Jan

doi:10.1007/978-3-031-35995-8_26

Cited by 6 publications

(2 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Modern machine learning systems require large amounts of data to perform satisfactorily. Although deep learning rapidly advances state-of-the-art results on a number of supervised learning tasks (Mukherjee and Awadallah, 2020; Kocoń and Maziarz, 2021;Srivastava et al, 2023;Koptyra et al, 2023), we can observe that gains strongly depend on large annotated datasets (Kanclerz et al, 2020;Shim et al, 2021;Kocoń et al, 2021b,a). Moreover, the data annotation process is very expensive (Kocoń et al, 2019a,b;Wierzba et al, 2021).…”

Section: Introductionmentioning

confidence: 78%

PALS: Personalized Active Learning for Subjective Tasks in NLP

Kanclerz,

Karanowski,

Bielaniewicz

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

For subjective NLP problems, such as classification of hate speech, aggression, or emotions, personalized solutions can be exploited. Then, the learned models infer about the perception of the content independently for each reader. To acquire training data, texts are commonly randomly assigned to users for annotation, which is expensive and highly inefficient. Therefore, for the first time, we suggest applying an active learning paradigm in a personalized context to better learn individual preferences. It aims to alleviate the labeling effort by selecting more relevant training samples. In this paper, we present novel Personalized Active Learning techniques for Subjective NLP tasks (PALS) to either reduce the cost of the annotation process or to boost the learning effect. Our five new measures allow us to determine the relevance of a text in the context of learning users' personal preferences. We validated them on three datasets: Wiki discussion texts individually labeled with aggression and toxicity, and on the Unhealthy Conversations dataset. Our PALS techniques outperform random selection even by more than 30%. They can also be used to reduce the number of necessary annotations while maintaining a given quality level. Personalized annotation assignments based on our controversy measure decrease the amount of data needed to just 25%-40% of the initial size.

show abstract

Section: Introductionmentioning

confidence: 78%

PALS: Personalized Active Learning for Subjective Tasks in NLP

Kanclerz,

Karanowski,

Bielaniewicz

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…While AI advancements, particularly in Natural Language Processing (NLP), have been groundbreaking, there remain significant chasms in their application in mental health, specifically regarding emotional interpretation. Several studies have been conducted, extensively highlighting the shortcomings of models like ChatGPT in nuanced emotional scenarios (Elyoseph et al, 2023;Koptyra et al, 2023). Even with state-of-the-art algorithms and vast datasets, the mechanical nature of these models prevents them from fully understanding or empathizing with the intricate labyrinth of human emotions.…”

Section: Limitations In Emotional Interpretationmentioning

confidence: 99%

Literature Review on the Double-Edged Sword of Ai in Mental Health: A Deep Dive Into Chatgpt's Capabilities and Limitations

Arjanto,

Senduk

2024

cmhp

View full text Add to dashboard Cite

Background: This paper focuses on the increasing relevance of AI in mental health care, particularly OpenAI's ChatGPT. It investigates the changing dynamics in mental health, analyzing ChatGPT's role, its benefits, drawbacks, and ethical complexities. Purpose: The objective is to assess ChatGPT's effectiveness in mental health, highlighting its strengths and limitations, and ethical issues. The study aims to understand how AI support can be balanced with the vital human aspect in mental health care. Methods: Comprehensive literature review of 7 pieces of literature from the Scopus database in 2023 (latest). Results: ChatGPT is found to be a useful initial mental health support tool, offering immediate access. However, it falls short in delivering the emotional depth that human health professionals provide. Key ethical concerns include data privacy and accountability. Conclusion: The study recommends a balanced approach, suggesting ChatGPT as an adjunct rather than a replacement for conventional mental health services. Effective use of ChatGPT in mental health care requires strict ethical guidelines and control measures to maintain the crucial human element in this field. Abstrak Latar Belakang: Artikel ini berfokus pada meningkatnya relevansi AI dalam layanan kesehatan mental, khususnya ChatGPT OpenAI. Laporan ini menyelidiki dinamika perubahan dalam kesehatan mental, menganalisis peran ChatGPT, manfaat, kelemahan, dan kompleksitas etikanya. Tujuan: Tujuannya adalah untuk menilai efektivitas ChatGPT dalam kesehatan mental, menyoroti kekuatan dan keterbatasannya, serta masalah etika. Studi ini bertujuan untuk memahami bagaimana dukungan AI dapat diseimbangkan dengan aspek vital manusia dalam perawatan kesehatan mental. Metode: Tinjauan pustaka yang menyeluruh terhadap 7 literatur dari database Scopus pada tahun 2023 (terbaru). Hasil: ChatGPT terbukti menjadi alat dukungan kesehatan mental awal yang berguna, menawarkan akses langsung. Namun, hal ini gagal dalam memberikan kedalaman emosional yang diberikan oleh para profesional kesehatan manusia. Masalah etika utama mencakup privasi dan akuntabilitas data. Kesimpulan: Studi ini merekomendasikan pendekatan yang seimbang, menyarankan ChatGPT sebagai tambahan dan bukan pengganti layanan kesehatan mental konvensional. Penggunaan ChatGPT yang efektif dalam perawatan kesehatan mental memerlukan pedoman etika yang ketat dan tindakan pengendalian untuk mempertahankan elemen manusia yang penting dalam bidang ini.

show abstract

A Web Approach for the Extraction, Analysis, and Visualization of Sentiments in Social Networks Regarding the Public Opinion on Politicians in Ecuador Using Natural Language Processing and High-Performance Computing Tools

Parra-Zambrano,

León-Paredes

2024

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

CLARIN-Emo: Training Emotion Recognition Models Using Human Annotation and ChatGPT

Cited by 6 publications

References 28 publications

PALS: Personalized Active Learning for Subjective Tasks in NLP

PALS: Personalized Active Learning for Subjective Tasks in NLP

Literature Review on the Double-Edged Sword of Ai in Mental Health: A Deep Dive Into Chatgpt's Capabilities and Limitations

A Web Approach for the Extraction, Analysis, and Visualization of Sentiments in Social Networks Regarding the Public Opinion on Politicians in Ecuador Using Natural Language Processing and High-Performance Computing Tools

Contact Info

Product

Resources

About