Publications

2026

Advances in Information Retrieval - 48th European Conference on Information Retrieval, ECIR 2026, Delft, The Netherlands, March 29 - April 2, 2026, Proceedings, Part I

Authors
Campos, R; Jatowt, A; Lan, Y; Aliannejadi, M; Bauer, C; MacAvaney, S; Anand, A; Ren, Z; Verberne, S; Bai, N; Mansoury, M;

Publication
ECIR (1)

Abstract

2026

Machine learning-based phonocardiographic analysis for pulmonary hypertension screening

Authors
Lobo, A; Almeida, MC; Costa, C; Oliveira, C; Gaudio, A; Giordano, N; Coimbra, M; Renna, F; Fontes-Carvalho, R;

Publication
EUROPEAN JOURNAL OF HEART FAILURE

Abstract

2026

SegNSP: Revisiting Next Sentence Prediction for Linear Text Segmentation

Authors
Isidro, J; Cunha, LF; Silvano, P; Jorge, A; Guimarães, N; Nunes, S; Campos, R;

Publication
CoRR

Abstract

2026

Proceedings of the 2026 Conference on Human Information Interaction and Retrieval, CHIIR 2026, Seattle, WA, USA, March 22-26, 2026

Authors
Shah, C; White, RW; Fourney, A; Lopes, CT; Trippas, JR;

Publication
CHIIR

Abstract

2026

Preface

Authors
Campos, R; Jatowt, A; Lan, Y; Aliannejadi, M; Bauer, C; MacAvaney, S; Anand, A; Ren, Z; Verberne, S; Bai, N; Mansoury, M;

Publication
Lecture Notes in Computer Science

Abstract
[No abstract available]

CloseRead Abstract

2026

Whisper-to-normal speech conversion using enhanced KNN-VC

Authors
Yamamura, CF; Scalassara, PR; Ferreira, A; Oliveira, MA;

Publication

Abstract
Whispers represent a common and secondary mechanism of communication. Nonetheless, individuals with aphonia, including those with laryngectomy, rely on whispers as their primary means of communication. Due to the substantial acoustic differences between whispered and normally phonated speech, the task of effective whispered-to-normal speech conversion (W2NSC) remains a significant challenge and has been widely discussed in the speech processing community. This study explores several enhancements to the k-nearest neighbors voice conversion (kNN-VC) model from multiple perspectives, including experiments with alternative feature extraction models, exploring fine-tuning of pre-trained models using Low-Rank Adaptation (LoRA), and the development of mapping strategies for parallel whispered and normal speech data using both KNN (pkNN-W2NSC) and multilayer perceptron (pMLP-W2NSC) approaches. Based on these experiments, a subjective evaluation using the MUSHRA test showed that the pMLP-W2NSC system achieved the highest overall performance among all evaluated configurations.

CloseRead Abstract

10
4555