Browsing by Author "Ansari, Luna"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Deep Learning for Automatic Classification of Speech Intensity Modes
    (2023-12-11) Ansari, Luna; Kadiri, Sudarsana; Perustieteiden korkeakoulu; Alku, Paavo
    One of the fundamental phenomena in speech processing is speech intensity. As a concept, speech intensity and its regulation help capture various aspects as well as changes in the mechanisms of the human speech production system. Except for a few studies, less is known about the automatic classification of vocal intensity categories. This study investigates speech intensity category classification by applying machine learning (ML) and deep learning (DL) classifiers in conjunction with different spectral features. A data set of speech recordings from 50 speakers reciting 25 sentences in four speech intensities (soft, normal, loud and very loud) has been analysed in this study. Specifically four spectral feature representations (static mel-frequency cepstral coefficients (MFCCs), dynamic MFCCs, spectrogram and mel-spectrogram) as well as their combinations are investigated. For the classifiers, three ML classifiers: Support Vector Machine (SVM), Random Forest (RF) and Adaboost, and four DL classifiers: Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and Bidirectional Long Short-Term Memory (BiLSTM) are explored. The best classification performance, accuracy of 76% is achieved with the combination of all features (dynamic MFCCs, spectrogram and mel-spectrogram) using the BiLSTM classifier.
  • No Thumbnail Available
    Item
    Ensemble Hybrid Learning Methods for Automated Depression Detection
    (2021-04-24) Ansari, Luna; Shaoxiong, Ji; Perustieteiden korkeakoulu; Käpylä, Maarit
  • Thumbnail Image
    Item
    Ensemble Hybrid Learning Methods for Automated Depression Detection
    (IEEE, 2023-02-01) Ansari, Luna; Ji, Shaoxiong; Chen, Qian; Cambria, Erik; Department of Computer Science; Professorship Marttinen P.; Department of Computer Science; Nanyang Technological University
    Changes in human lifestyle have led to an increase in the number of people suffering from depression over the past century. Although in recent years, rates of diagnosing mental illness have improved, many cases remain undetected. Automated detection methods can help identify depressed or individuals at risk. An understanding of depression detection requires effective feature representation and analysis of language use. In this article, text classifiers are trained for depression detection. The key objective is to improve depression detection performance by examining and comparing two sets of methods: hybrid and ensemble. The results show that ensemble models outperform the hybrid model classification results. The strength and effectiveness of the combined features demonstrate that better performance can be achieved by multiple feature combinations and proper feature selection.
  • Thumbnail Image
    Item
    MentalBERT: Publicly Available Pretrained Language Models for Mental Healthcare
    (2022) Ji, Shaoxiong; Zhang, Tianlin; Ansari, Luna; Fu, Jie; Tiwari, Prayag; Cambria, Erik; Department of Computer Science; University of Manchester; Aalto University; Mila - Quebec Artificial Intelligence Institute; Professorship Marttinen P.; Nanyang Technological University; Calzolari, N; Bechet, F; Blache, P; Choukri, K; Cieri, C; Declerck, T; Goggi, S; Isahara, H; Maegaard, B; Mazo, H; Odijk, H; Piperidis, S
    Mental health is a critical issue in modern society, and mental disorders could sometimes turn to suicidal ideation without adequate treatment. Early detection of mental disorders and suicidal ideation from social content provides a potential way for effective social intervention. Recent advances in pretrained contextualized language representations have promoted the development of several domain-specific pretrained models and facilitated several downstream applications. However, there are no existing pretrained language models for mental healthcare. This paper trains and releases two pretrained masked language models, i.e., MentalBERT and MentalRoBERTa, to benefit machine learning for the mental healthcare research community. Besides, we evaluate our trained domain-specific models and several variants of pretrained language models on several mental disorder detection benchmarks and demonstrate that language representations pretrained in the target domain improve the performance of mental health detection tasks.
  • Thumbnail Image
    Item
    On micro foundations of corporate sustainability
    (Aalto University, 2021) Ansari, Luna; Johtamisen laitos; Department of Management Studies; Kauppakorkeakoulu; School of Business; Granqvist, Nina, Assoc. Prof., Aalto University, Department of Management Studies, Finland; Vaara, Eero, Prof., Oxford University, UK
    One of the realms where contradictions manifest in organizations is corporate sustainability, in which managers confront tensions between complex and interconnected concerns for the natural environment, social welfare, and economic prosperity. In this doctoral dissertation, I explore what explains why actors and organizations differ in their understanding of the tension between the three pillars – social, environmental, and economical – of corporate sustainability. This question is approached by considering the affective as well as discursive dynamics of dealing with corporate sustainability. This is a study about the lived experience of the challenges of corporate sustainability. The first essay characterizes the reconciliation of diverging goals in sustainability from a discursive point. The starting point of this analysis is that a central part of reconciliation takes place through language. The second essay focuses on the affective underpinning of living with tensions of sustainability. In particular, I argue that in the sustainability management context, self-regulation supports managers' capacity to accept and promote new environmental frames. In the third essay, I empirically analyze collective action taking and apply media text to study group-based affect regulation. Through these essays, I contribute to the understanding of the challenges that individuals face in terms of adapting and adjusting to new ways of considering corporate sustainability. By analyzing the working and living realities of various groups of actors, this study provides a holistic understanding of coping mechanisms in mobilizing responses to corporate sustainability demands. This study reveals the affective dimensions individuals face in their endeavors to take on, promote, and institutionalize new perspectives of looking at the phenomenon of corporate sustainability.