arXiv subjects

Maria Luisa Gorno-Tempini

Publications and source records attributed to Maria Luisa Gorno-Tempini.

3 recordsLinked to original sources

arXiv2026

K-Function: Joint Pronunciation Transcription and Feedback for Evaluating Kids Language Function

Evaluating young children's language is challenging for automatic speech recognizers due to high-pitched voices, prolonged sounds, and limited data. We introduce K-Function, a framework that combines accurate sub-word transcription with objective, Large Language Model (LLM)-driven scoring. Its core, Kids-Weighted Finite State Transducer (K-WFST), merges an acoustic phoneme encoder with a phoneme-similarity model to capture child-specific speech errors while remaining fully interpretable. K-WFST achieves a 1.39 % phoneme error rate on MyST and 8.61 % on Multitudes-an absolute improvement of 10.47 % and 7.06 % over a greedy-search decoder. These high-quality transcripts are used by an LLM to grade verbal skills, developmental milestones, reading, and comprehension, with results that align closely with human evaluators. Our findings show that precise phoneme recognition is essential for creating an effective assessment framework, enabling scalable language screening for children.

cs.CL↗

arXiv2025

LCS-CTC: Leveraging Soft Alignments to Enhance Phonetic Transcription Robustness

Phonetic speech transcription is crucial for fine-grained linguistic analysis and downstream speech applications. While Connectionist Temporal Classification (CTC) is a widely used approach for such tasks due to its efficiency, it often falls short in recognition performance, especially under unclear and nonfluent speech. In this work, we propose LCS-CTC, a two-stage framework for phoneme-level speech recognition that combines a similarity-aware local alignment algorithm with a constrained CTC training objective. By predicting fine-grained frame-phoneme cost matrices and applying a modified Longest Common Subsequence (LCS) algorithm, our method identifies high-confidence alignment zones which are used to constrain the CTC decoding path space, thereby reducing overfitting and improving generalization ability, which enables both robust recognition and text-free forced alignment. Experiments on both LibriSpeech and PPA demonstrate that LCS-CTC consistently outperforms vanilla CTC baselines, suggesting its potential to unify phoneme modeling across fluent and non-fluent speech.

eess.AS↗

arXiv2025

Multi-object Data Integration in the Study of Primary Progressive Aphasia

This article focuses on a multi-modal imaging data application where structural/anatomical information from gray matter (GM) and brain connectivity information in the form of a brain connectome network from functional magnetic resonance imaging (fMRI) are available for a number of subjects with different degrees of primary progressive aphasia (PPA), a neurodegenerative disorder (ND) measured through a speech rate measure on motor speech loss. The clinical/scientific goal in this study becomes the identification of brain regions of interest significantly related to the speech rate measure to gain insight into ND patterns. Viewing the brain connectome network and GM images as objects, we develop an integrated object response regression framework of network and GM images on the speech rate measure. A novel integrated prior formulation is proposed on network and structural image coefficients in order to exploit network information of the brain connectome while leveraging the interconnections among the two objects. The principled Bayesian framework allows the characterization of uncertainty in ascertaining a region being actively related to the speech rate measure. Our framework yields new insights into the relationship of brain regions associated with PPA, offering a deeper understanding of neuro-degenerative patterns of PPA. The supplementary file adds details about posterior computation and additional empirical results.

q-bio.NC↗