Toshevska, Martina

Preferred name

Toshevska, Martina

Official Name

Toshevska, Martina

Main Affiliation

Now showing 1 - 10 of 19

Some of the metrics are blocked by your
consent settings
Explorations into Deep Learning Text Architectures for Dense Image Captioning
(2020-09)
;
Stojanovska, Frosina
;
;
;
Some of the metrics are blocked by your
consent settings
LLM-Based Text Style Transfer: Have We Taken a Step Forward?
(Institute of Electrical and Electronics Engineers (IEEE), 2025-03-06)
;
Some of the metrics are blocked by your
consent settings
Emotion-Aware Teaching Robot: Learning to Adjust to User’s Emotional State
(Springer, Cham, 2018-09-17)
Stojanovska, Frosina
;
;
;
Robots today are taking more and more complex roles thus they are getting smarter and more human-like. One complex function, specific to social robots, is the role of robots in human-robot interaction. They are helpful in the process of social human-robot interaction while performing a specific task like teaching, assisting, entertaining, etc. The ability to recognize emotions has a significant role for social robots. A robot that can understand emotions could be able to interact according to that emotion. In this paper, we propose a model for robotic behavior adapting to the user’s emotions. The humanoid robot Nao is used in the role of emotion-aware teacher for teaching math. Its main purpose is to teach and entertain the user while adapting its behavior to the user’s emotional state derived from the facial expression. The robot uses reinforcement learning to learn which action to perform in a specific emotional state. It employs the Q-learning algorithm, maximizing the next action’s award - a value that depends on the current emotional state of the user. An experimental study with a selected group of subjects is conducted to assess the proposed behavior. We evaluated the robot’s ability to recognize emotions and the subjects’ experience of interacting with the robot.
Some of the metrics are blocked by your
consent settings
Explorations into deep neural models for emotion recognition
(Springer, Cham, 2018-09-17)
Stojanovska, Frosina
;
;
Deep emotion recognition is the central objective of our recent research efforts. This study examines the capability of several deep learning architectures and word embeddings to classify emotions on two Twitter datasets. We have identified several aspects worth investigating that appeared to challenge and contrast previously established notion that semantic information is captured by distributional word representations. Our evidence has shown that extending the word embeddings to account for the use of emojis and incorporating a suitable lexicon of emotional words can lead to a better classification of the emotional content carried by Twitter messages.
Some of the metrics are blocked by your
consent settings
Question Answering with Deep Learning: A Survey
(2019)
;
;
Some of the metrics are blocked by your
consent settings
Exploring Large Language Models for Data Augmentation: A Case Study for Text Style Transfer
(IEEE, 2025-06-02)
;
;
Text style transfer is the task that involves modifying a sentence to adapt to a desired target style while preserving its original meaning. It often requires high-quality parallel datasets that are not always available. This paper explores data augmentation techniques for text style transfer, leveraging large language models (LLMs) to address the challenge of dataset scarcity. Our approach generates synthetic parallel data by prompting LLMs to paraphrase and/or rewrite sentences in diverse styles, enabling the creation of larger and more varied datasets. We demonstrate the applicability of this approach across three tasks: formality transfer with the GYAFC dataset, sentiment transfer with the Yelp dataset, and personal style transfer with the Shakespeare dataset. This work introduces an approach to enhance dataset availability, aiming to foster further research in the field and support a broader application of LLMs. The experiments were performed only with English language datasets.
Some of the metrics are blocked by your
consent settings
Hate speech detection using XGBoost, CNN+LSTM and BERT
(2023)
Zdravkovska, Jovana
;
Petrushevska, Magdalena
;
Paunkoska, Sara
;
;
Some of the metrics are blocked by your
consent settings
Fake News Detection - An Overview of Methods and Datasets
(2023)
Graorkovski, Goran
;
;
Some of the metrics are blocked by your
consent settings
A Review of Text Style Transfer using Deep Learning
(IEEE, 2021-09-28)
;
Style is an integral component of a sentence indicated by the choice of words a person makes. Different people have different ways of expressing themselves, however, they adjust their speaking and writing style to a social context, an audience, an interlocutor or the formality of an occasion. Text style transfer is defined as a task of adapting and/or changing the stylistic manner in which a sentence is written, while preserving the meaning of the original sentence. A systematic review of text style transfer methodologies using deep learning is presented in this paper. We point out the technological advances in deep neural networks that have been the driving force behind current successes in the fields of natural language understanding and generation. The review is structured around two key stages in the text style transfer process, namely, representation learning and sentence generation in a new style. The discussion highlights the commonalities and differences between proposed solutions as well as challenges and opportunities that are expected to direct and foster further research in the field.
Some of the metrics are blocked by your
consent settings
Enhancing LLMs with LoRA Fine-Tuning Using Medical Data and Knowledge Graph Enrichment for Improved Healthcare Outcomes
(IEEE, 2025-06-02)
Jankov, A.
;
;
This research paper investigates the enhancement of large language models (LLMs) within the medical domain, focusing on members of the Llama family of LLMs. While LLMs have demonstrated remarkable success across various general-purpose natural language processing tasks, their application in specialized domains like medicine is often hindered by limited training on domain-specific data, resulting in suboptimal accuracy and contextual relevance. To address these limitations, this research employs low-rank adaptation (LoRA) to fine-tune LLMs on real-world patientphysician dialogues, effectively capturing the intricacies of medical discourse. Additionally, the knowledge of the LLM is enriched with the SPOKE knowledge graph, a structured repository of medical domain information, allowing the model to generate outputs that are both contextually and scientifically grounded. The experimental results underscore the transformative impact of this dual approach, demonstrating significant advancements in tasks such as automatic diagnosis generation and personalized drug recommendation. However, this research should be viewed as an exploratory proof of concept. Significant limitations, including the constrained evaluation scope and the critical need for expert clinical validation and thorough ethical review, must be addressed in future work before considering real-world applicability.

Toshevska, Martina

Filters

Settings

Sort By

Results per page