arXiv subjects

Soumya Dutta

Publications and source records attributed to Soumya Dutta.

At least 19 recordsLinked to original sources

arXiv2026

On Improving Faithfulness of Podcasts from Documents

Large language models (LLMs) are increasingly used to generate long-form conversational content such as podcasts from textual sources. While these systems produce fluent and engaging narratives, they often introduce ungrounded information. In this work, we present the first systematic study of faithfulness in document-grounded podcast generation, where grounding must be maintained across conversational turns in long-form, multi-speaker transcripts. We construct a dataset of over 1500 documents spanning five domains and generate podcast transcripts using multiple LLMs. We introduce a turn-level LLM-as-a-judge framework for evaluating whether conversational turns are supported by the source document, and validate its reliability through human studies. Our analysis shows that even state-of-the-art models, including GPT-4o, frequently generate ungrounded content. To mitigate this issue, we propose catch-n-repair, a model-agnostic framework that detects and rewrites unfaithful conversational turns while preserving conversational flow. Experiments demonstrate consistent improvements in faithfulness across both in-domain and out-of-domain settings.

cs.CL↗

arXiv2026

SAiW: Source-Attributable Invisible Watermarking for Proactive Deepfake Defense

Deepfakes generated by modern generative models pose a serious threat to information integrity, digital identity, and public trust. Existing detection methods are largely reactive, attempting to identify manipulations after they occur and often failing to generalize across evolving generation techniques. This motivates the need for proactive mechanisms that secure media authenticity at the time of creation. In this work, we introduce SAiW, a Source-Attributed Invisible watermarking Framework for proactive deepfake defense and media provenance verification. Unlike conventional watermarking methods that treat watermark payloads as generic signals, SAiW formulates watermark embedding as a source-conditioned representation learning problem, where watermark identity encodes the originating source and modulates the embedding process to produce discriminative and traceable signatures. The framework integrates feature-wise linear modulation to inject source identity into the embedding network, enabling scalable multi-source watermark generation. A perceptual guidance module derived from human visual system priors ensures that watermark perturbations remain visually imperceptible while maintaining robustness. In addition, a dual-purpose forensic decoder simultaneously reconstructs the embedded watermark and performs source attribution, providing both automated verification and interpretable forensic evidence. Extensive experiments across multiple deepfake datasets demonstrate that SAiW achieves high perceptual quality while maintaining strong robustness against compression, filtering, noise, geometric transformations, and adversarial perturbations. By binding digital media to its origin through invisible yet verifiable markers, SAiW enables reliable authentication and source attribution, providing a scalable foundation for proactive deepfake defense and trustworthy media provenance.

cs.AI↗

arXiv2026

Textless and Non-Parallel Speech-to-Speech Emotion Style Transfer

Given a pair of source and reference speech recordings, speech-to-speech (S2S) emotion style transfer involves the generation of an output speech that mimics the emotion characteristics of the reference while preserving the content and speaker attributes of the source. In this paper, we propose a speech-to-speech zero-shot emotion style transfer framework, termed S2S Zero-shot Emotion Style Transfer (S2S-ZEST), that enables the transfer of emotional attributes from the reference to the source while retaining the speaker identity and speech content. The S2S-ZEST framework consists of an analysis-synthesis pipeline in which the analysis module extracts semantic tokens, speaker representations, and emotion embeddings from speech. Using these representations, a pitch contour estimator and a duration predictor are learned. Further, a synthesis module is designed to generate speech based on the input representations and the derived factors. The analysis-synthesis pipeline is trained using an auto-encoding objective to enable efficient resynthesis during inference. For S2S emotion style transfer, the emotion embedding extracted from the reference speech along with the remaining representations from the source speech are used in the synthesis module to generate the style-transferred speech. In our experiments, we evaluate the converted speech on content and speaker preservation (with respect to the source) as well as on the effectiveness of the emotion style transfer (with respect to the reference). The proposed framework demonstrates improved emotion style transfer performance over prior methods in a textless and non-parallel setting. We also illustrate the application of the proposed work for data augmentation in emotion recognition tasks.

eess.AS↗

arXiv2026

Acoustic and Semantic Modeling of Emotion in Spoken Language

Emotions play a central role in human communication, shaping trust, engagement, and social interaction. As artificial intelligence systems powered by large language models become increasingly integrated into everyday life, enabling them to reliably understand and generate human emotions remains an important challenge. While emotional expression is inherently multimodal, this thesis focuses on emotions conveyed through spoken language and investigates how acoustic and semantic information can be jointly modeled to advance both emotion understanding and emotion synthesis from speech. The first part of the thesis studies emotion-aware representation learning through pre-training. We propose strategies that incorporate acoustic and semantic supervision to learn representations that better capture affective cues in speech. A speech-driven supervised pre-training framework is also introduced to enable large-scale emotion-aware text modeling without requiring manually annotated text corpora. The second part addresses emotion recognition in conversational settings. Hierarchical architectures combining cross-modal attention and mixture-of-experts fusion are developed to integrate acoustic and semantic information across conversational turns. Finally, the thesis introduces a textless and non-parallel speech-to-speech framework for emotion style transfer that enables controllable emotional transformations while preserving speaker identity and linguistic content. The results demonstrate improved emotion transfer and show that style-transferred speech can be used for data augmentation to improve emotion recognition.

eess.AS↗

arXiv2026

A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations

Emotion Recognition in Conversations (ERC) presents unique challenges, requiring models to capture the temporal flow of multi-turn dialogues and to effectively integrate cues from multiple modalities. We propose Mixture of Speech-Text Experts for Recognition of Emotions (MiSTER-E), a modular Mixture-of-Experts (MoE) framework designed to decouple two core challenges in ERC: modality-specific context modeling and multimodal information fusion. MiSTER-E leverages large language models (LLMs) fine-tuned for both speech and text to provide rich utterance-level embeddings, which are then enhanced through a convolutional-recurrent context modeling layer. The system integrates predictions from three experts-speech-only, text-only, and cross-modal-using a learned gating mechanism that dynamically weighs their outputs. To further encourage consistency and alignment across modalities, we introduce a supervised contrastive loss between paired speech-text representations and a KL-divergence-based regulariza-tion across expert predictions. Importantly, MiSTER-E does not rely on speaker identity at any stage. Experiments on three benchmark datasets-IEMOCAP, MELD, and MOSI-show that our proposal achieves 70.9%, 69.5%, and 87.9% weighted F1-scores respectively, outperforming several baseline speech-text ERC systems. We also provide various ablations to highlight the contributions made in the proposed approach.

cs.CL↗

arXiv2026

REV-INR: Regularized Evidential Implicit Neural Representation for Uncertainty-Aware Volume Visualization

Applications of Implicit Neural Representations (INRs) have emerged as a promising deep learning approach for compactly representing large volumetric datasets. These models can act as surrogates for volume data, enabling efficient storage and on-demand reconstruction via model predictions. However, conventional deterministic INRs only provide value predictions without insights into the model's prediction uncertainty or the impact of inherent noisiness in the data. This limitation can lead to unreliable data interpretation and visualization due to prediction inaccuracies in the reconstructed volume. Identifying erroneous results extracted from model-predicted data may be infeasible, as raw data may be unavailable due to its large size. To address this challenge, we introduce REV-INR, Regularized Evidential Implicit Neural Representation, which learns to predict data values accurately along with the associated coordinate-level data uncertainty and model uncertainty using only a single forward pass of the trained REV-INR during inference. By comprehensively comparing and contrasting REV-INR with existing well-established deep uncertainty estimation methods, we show that REV-INR achieves the best volume reconstruction quality with robust data (aleatoric) and model (epistemic) uncertainty estimates using the fastest inference time. Consequently, we demonstrate that REV-INR facilitates assessment of the reliability and trustworthiness of the extracted isosurfaces and volume visualization results, enabling analyses to be solely driven by model-predicted data.

cs.LG↗

arXiv2025

Compressive Modeling and Visualization of Multivariate Scientific Data using Implicit Neural Representation

The extensive adoption of Deep Neural Networks has led to their increased utilization in challenging scientific visualization tasks. Recent advancements in building compressed data models using implicit neural representations have shown promising results for tasks like spatiotemporal volume visualization and super-resolution. Inspired by these successes, we develop compressed neural representations for multivariate datasets containing tens to hundreds of variables. Our approach utilizes a single network to learn representations for all data variables simultaneously through parameter sharing. This allows us to achieve state-of-the-art data compression. Through comprehensive evaluations, we demonstrate superior performance in terms of reconstructed data quality, rendering and visualization quality, preservation of dependency information among variables, and storage efficiency.

cs.LG↗

arXiv2025

Leveraging Content and Acoustic Representations for Speech Emotion Recognition

Speech emotion recognition (SER), the task of identifying the expression of emotion from spoken content, is challenging due to the difficulty in extracting representations that capture emotional attributes from speech. The scarcity of labeled datasets further complicates the challenge where large models are prone to over-fitting. In this paper, we propose CARE (Content and Acoustic Representations of Emotions), where we design a dual encoding scheme which emphasizes semantic and acoustic factors of speech. While the semantic encoder is trained using distillation from utterance-level text representations, the acoustic encoder is trained to predict low-level frame-wise features of the speech signal. The proposed dual encoding scheme is a base-sized model trained only on unsupervised raw speech. With a simple light-weight classification model trained on the downstream task, we show that the CARE embeddings provide effective emotion recognition on a variety of datasets. We compare the proposal with several other self-supervised models as well as recent large-language model based approaches. In these evaluations, the proposed CARE is shown to be the best performing model based on average performance across 8 diverse datasets. We also conduct several ablation studies to analyze the importance of various design choices.

eess.AS↗

arXiv2025

ABHINAYA -- A System for Speech Emotion Recognition In Naturalistic Conditions Challenge

Speech emotion recognition (SER) in naturalistic settings remains a challenge due to the intrinsic variability, diverse recording conditions, and class imbalance. As participants in the Interspeech Naturalistic SER Challenge which focused on these complexities, we present Abhinaya, a system integrating speech-based, text-based, and speech-text models. Our approach fine-tunes self-supervised and speech large language models (SLLM) for speech representations, leverages large language models (LLM) for textual context, and employs speech-text modeling with an SLLM to capture nuanced emotional cues. To combat class imbalance, we apply tailored loss functions and generate categorical decisions through majority voting. Despite one model not being fully trained, the Abhinaya system ranked 4th among 166 submissions. Upon completion of training, it achieved state-of-the-art performance among published results, demonstrating the effectiveness of our approach for SER in real-world conditions.

cs.SD↗

arXiv2025

LLM supervised Pre-training for Multimodal Emotion Recognition in Conversations

Emotion recognition in conversations (ERC) is challenging due to the multimodal nature of the emotion expression. In this paper, we propose to pretrain a text-based recognition model from unsupervised speech transcripts with LLM guidance. These transcriptions are obtained from a raw speech dataset with a pre-trained ASR system. A text LLM model is queried to provide pseudo-labels for these transcripts, and these pseudo-labeled transcripts are subsequently used for learning an utterance level text-based emotion recognition model. We use the utterance level text embeddings for emotion recognition in conversations along with speech embeddings obtained from a recently proposed pre-trained model. A hierarchical way of training the speech-text model is proposed, keeping in mind the conversational nature of the dataset. We perform experiments on three established datasets, namely, IEMOCAP, MELD, and CMU- MOSI, where we illustrate that the proposed model improves over other benchmarks and achieves state-of-the-art results on two out of these three datasets.

eess.AS↗

arXiv2024

Uncertainty-Informed Volume Visualization using Implicit Neural Representation

The increasing adoption of Deep Neural Networks (DNNs) has led to their application in many challenging scientific visualization tasks. While advanced DNNs offer impressive generalization capabilities, understanding factors such as model prediction quality, robustness, and uncertainty is crucial. These insights can enable domain scientists to make informed decisions about their data. However, DNNs inherently lack ability to estimate prediction uncertainty, necessitating new research to construct robust uncertainty-aware visualization techniques tailored for various visualization tasks. In this work, we propose uncertainty-aware implicit neural representations to model scalar field data sets effectively and comprehensively study the efficacy and benefits of estimated uncertainty information for volume visualization tasks. We evaluate the effectiveness of two principled deep uncertainty estimation techniques: (1) Deep Ensemble and (2) Monte Carlo Dropout (MCDropout). These techniques enable uncertainty-informed volume visualization in scalar field data sets. Our extensive exploration across multiple data sets demonstrates that uncertainty-aware models produce informative volume visualization results. Moreover, integrating prediction uncertainty enhances the trustworthiness of our DNN model, making it suitable for robustly analyzing and visualizing real-world scientific volumetric data sets.

cs.GR↗

arXiv2024

Uncertainty-Aware Deep Neural Representations for Visual Analysis of Vector Field Data

The widespread use of Deep Neural Networks (DNNs) has recently resulted in their application to challenging scientific visualization tasks. While advanced DNNs demonstrate impressive generalization abilities, understanding factors like prediction quality, confidence, robustness, and uncertainty is crucial. These insights aid application scientists in making informed decisions. However, DNNs lack inherent mechanisms to measure prediction uncertainty, prompting the creation of distinct frameworks for constructing robust uncertainty-aware models tailored to various visualization tasks. In this work, we develop uncertainty-aware implicit neural representations to model steady-state vector fields effectively. We comprehensively evaluate the efficacy of two principled deep uncertainty estimation techniques: (1) Deep Ensemble and (2) Monte Carlo Dropout, aimed at enabling uncertainty-informed visual analysis of features within steady vector field data. Our detailed exploration using several vector data sets indicate that uncertainty-aware models generate informative visualization results of vector field features. Furthermore, incorporating prediction uncertainty improves the resilience and interpretability of our DNN model, rendering it applicable for the analysis of non-trivial vector field data sets.

cs.GR↗

arXiv2024

Visual Analysis of Prediction Uncertainty in Neural Networks for Deep Image Synthesis

Ubiquitous applications of Deep neural networks (DNNs) in different artificial intelligence systems have led to their adoption in solving challenging visualization problems in recent years. While sophisticated DNNs offer an impressive generalization, it is imperative to comprehend the quality, confidence, robustness, and uncertainty associated with their prediction. A thorough understanding of these quantities produces actionable insights that help application scientists make informed decisions. Unfortunately, the intrinsic design principles of the DNNs cannot beget prediction uncertainty, necessitating separate formulations for robust uncertainty-aware models for diverse visualization applications. To that end, this contribution demonstrates how the prediction uncertainty and sensitivity of DNNs can be estimated efficiently using various methods and then interactively compared and contrasted for deep image synthesis tasks. Our inspection suggests that uncertainty-aware deep visualization models generate illustrations of informative and superior quality and diversity. Furthermore, prediction uncertainty improves the robustness and interpretability of deep visualization models, making them practical and convenient for various scientific domains that thrive on visual analyses.

cs.CV↗

arXiv2024

HCAM -- Hierarchical Cross Attention Model for Multi-modal Emotion Recognition

Emotion recognition in conversations is challenging due to the multi-modal nature of the emotion expression. We propose a hierarchical cross-attention model (HCAM) approach to multi-modal emotion recognition using a combination of recurrent and co-attention neural network models. The input to the model consists of two modalities, i) audio data, processed through a learnable wav2vec approach and, ii) text data represented using a bidirectional encoder representations from transformers (BERT) model. The audio and text representations are processed using a set of bi-directional recurrent neural network layers with self-attention that converts each utterance in a given conversation to a fixed dimensional embedding. In order to incorporate contextual knowledge and the information across the two modalities, the audio and text embeddings are combined using a co-attention layer that attempts to weigh the utterance level embeddings relevant to the task of emotion recognition. The neural network parameters in the audio layers, text layers as well as the multi-modal co-attention layers, are hierarchically trained for the emotion classification task. We perform experiments on three established datasets namely, IEMOCAP, MELD and CMU-MOSI, where we illustrate that the proposed model improves significantly over other benchmarks and helps achieve state-of-art results on all these datasets.

eess.AS↗

arXiv2024

Zero Shot Audio to Audio Emotion Transfer With Speaker Disentanglement

The problem of audio-to-audio (A2A) style transfer involves replacing the style features of the source audio with those from the target audio while preserving the content related attributes of the source audio. In this paper, we propose an efficient approach, termed as Zero-shot Emotion Style Transfer (ZEST), that allows the transfer of emotional content present in the given source audio with the one embedded in the target audio while retaining the speaker and speech content from the source. The proposed system builds upon decomposing speech into semantic tokens, speaker representations and emotion embeddings. Using these factors, we propose a framework to reconstruct the pitch contour of the given speech signal and train a decoder that reconstructs the speech signal. The model is trained using a self-supervision based reconstruction loss. During conversion, the emotion embedding is alone derived from the target audio, while rest of the factors are derived from the source audio. In our experiments, we show that, even without using parallel training data or labels from the source or target audio, we illustrate zero shot emotion transfer capabilities of the proposed ZEST model using objective and subjective quality evaluations.

eess.AS↗

arXiv2023

Dynamic Spatio-Temporal Summarization using Information Based Fusion

In the era of burgeoning data generation, managing and storing large-scale time-varying datasets poses significant challenges. With the rise of supercomputing capabilities, the volume of data produced has soared, intensifying storage and I/O overheads. To address this issue, we propose a dynamic spatio-temporal data summarization technique that identifies informative features in key timesteps and fuses less informative ones. This approach minimizes storage requirements while preserving data dynamics. Unlike existing methods, our method retains both raw and summarized timesteps, ensuring a comprehensive view of information changes over time. We utilize information-theoretic measures to guide the fusion process, resulting in a visual representation that captures essential data patterns. We demonstrate the versatility of our technique across diverse datasets, encompassing particle-based flow simulations, security and surveillance applications, and biological cell interactions within the immune system. Our research significantly contributes to the realm of data management, introducing enhanced efficiency and deeper insights across diverse multidisciplinary domains. We provide a streamlined approach for handling massive datasets that can be applied to in situ analysis as well as post hoc analysis. This not only addresses the escalating challenges of data storage and I/O overheads but also unlocks the potential for informed decision-making. Our method empowers researchers and experts to explore essential temporal dynamics while minimizing storage requirements, thereby fostering a more effective and intuitive understanding of complex data behaviors.

cs.CV↗

arXiv2023

Giant conductance of PSS:PEDOT micro-surfaces induced by microbubble lithography

We provide direct evidence of the effects of interface engineering of various substrates by Microbubble lithography (MBL). We choose a model organic plastic (or polymer) poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS), with conductivity of 140 S/cm, as a representative organic system to showcase our technique. Thus, we fabricate permanent patterns of PEDOT:PSS on glass, followed by a flexible PDMS substrate, and observe conductivity enhancement of 5 times on the former (694 S/cm), and 20 times (2844 S/cm) on the latter, without the use of external doping agents or invasive chemical treatment. Probing the patterned interface, we observe that MBL is able to tune the conformational states of PEDOT:PSS from coils in the pristine form, to extended coils on glass, and almost linear structures in PDMS due to its more malleable liquid-like interface. This results in higher ordering and vanishing grain boundaries leading to the highest conductivity of PEDOT:PSS on PDMS substrates.

physics.chem-ph↗

arXiv2022

Nature of excitons in PPDT2FBT: PCBM solar cell: Role played by PCBM

In organic semiconductor based bulk heterojunction solar cells, the presence of acceptor increases the formation of charge transfer (CT) excitons, thereby leading to higher exciton dissociation probabilities. In this work we used steady state EA measurements to probe the change in the nature of excitons as the blend composition of the solar cell active layer material is varied. We investigated blends of poly[(2,5-bis(2-hexyldecyloxy)phenylene)-alt-(5,6-difluoro-4,7-di(thiophen-2-yl)benzo[c]-[1,2,5]thiadiazole)] (PPDT2FBT) and (6,6)-Phenyl C71 butyric acid methyl ester (PCBM). Analysis of the EA spectra showed that in presence of fullerene based acceptor, like PCBM, CT characteristics of the excitons were modified, though, no new CT signature was observed in the blend. Enhancement in the CT characteristic in the blend was reflected in the photoluminescence (PL) measurements of the blends, where, PL quenching of $\sim$ 63\% was observed for 1\% PCBM. The quenching reaches saturation at about 20\% PCBM. However, the maximum efficiency of the devices was obtained for the blend having 50\% PCBM. Comparing experimental results with simulations, the variation of the device efficiency with PCBM percentage was shown to be arising from multiple factors like increase in polarizability and dipole moment of excitons, and the efficiency of the carrier collection from the bulk of the active layer.

physics.app-ph↗