Loquens

Evaluation of German Automatic Speech Recognition solutions in the context of speech and language therapy support of people with aphasia

2024-07-04T22:17:11+02:00

Those who suffer from aphasia benefit from digital speech and language therapy solutions, and automatic speech recognition (ASR) has been already used for giving feedback on the correctness of the answers in naming exercises. AphaDIGITAL application is to provide German-speaking users with detailed feedback on phonemic/phonetic and semantic errors, based on automatic speech and language processing. For this purpose, open-source ASR solutions for German were evaluated on different corpora of atypical speech, including two small datasets with aphasic speech samples. Character error rate, the number of precisely recognized items and empty outputs served as evaluation metrics. The four selected models are generally robust to the deteriorated condition of speech and audio quality and consistently outperform commercial models in atypical speech recognition. Applying error acceptance threshold, additional use of phonemic error rate, and other valuable insights for ASR implementation in aphaDIGITAL are discussed.

Russian spoken by Kazakh young adults: a phonetic and phonological analysis

2024-05-13T12:20:23+02:00

This study examines the phonetics and phonology of Russian spoken by young adults from different regions of Kazakhstan, focusing on stops, fricatives, and affricates absent in Kazakh but present in Russian. The analysis highlights the variability in phoneme production based on regional origins and exposure to Russian. Some participants showed patterns similar to those of Russian native speakers, while others exhibited more variability, including phenomena not typically found in Russian. The study also found a de-palatalization phenomenon in post-alveolar fricatives, suggesting a loss of phonological distinction. These findings provide insights into the linguistic adaptation of Kazakh speakers, revealing how they navigate and assimilate Russian phonological features.

Invading the hearer’s auditory space. A prosodic analysis of Italian insults

2024-09-09T10:49:27+02:00

Impoliteness is a broad field of research that also encompasses insults, which are forms of verbal attack that involve a degree of injury and social censure. From a linguistic standpoint, insults are characterised by distinctive verbal and non-verbal cues, including specific phonetic and prosodic alterations. The aim of this research was to analyse, for the first time, the prosody of Italian insults using a spectro-acoustic approach. The study utilised data from 15 oral role plays performed by six professional theatre actors. The objective of the study was to investigate the extent to which duration, intensity, and pitch range are employed to convey the specific meaning of an invasion of the listener’s auditory space. The analysis demonstrated that insults exhibited longer nuclear vowels and higher intensity values than the control speech sample. Conversely, the pitch range of the target utterances was narrowed, with the majority of f₀ values situated within the high-frequency range.

The melodic profile of declarative and interrogative utterances of Swedish speakers of Spanish

2024-11-11T05:25:29+01:00

Declarative and interrogative utterances in Spanish spoken by Swedish tend to be produced as suspended and generally lack of the melodic variation of Spanish. The aim of this paper is to describe and compare their melodic features to determine if they are produced in different ways to differentiate their communicative function in the interlanguage. We analyse a corpus of 180 utterances (108 declaratives and 74 absolute interrogatives) from the spontaneous speech of 45 native speakers of Swedish and of Spanish FL with intermediate level, using the Melodic Analysis of Speech (MAS) method. The results show that there are significant differences especially in the final inflection, with more flat or descending endings in declaratives and more ascending endings in interrogatives; although in both cases they are produced with melodic features of suspended intonation in Spanish. More marked increases are also detected in the body tonal inflections of the interrogatives.

Rhythmic and frequency prosody in clinical linguistics: in search of phonetic biomarkers for Alzheimer’s disease and Major Depression

2024-10-12T13:01:24+02:00

This paper delves into the prosodic features of speech affected in AD and MD, the type of pathology they are associated with (aphasia, apraxia or aprosodia), as well as the contribution of these alterations to the neurolinguistic field (hemispheric specialization).We used a corpus of spontaneous speech from patients with AD (n=10) and MD (n=10), and a control group with HS (n=10). The results showed a discriminatory significance of VnPVI (in MD), and (Δ)f0 and %V; furthermore, %V is a strong variable to distinguish between speakers with and without pathology, as well as to discriminate between AD and MD along with other variables. There is, however, no correlation with verbal fluency (semantic and phonological) or with MMSE, which would imply a motor type deficit: apraxic type in AD and aprosodic type in MD. These results show certain phonological and typological conditioning, among other issues.

Vowel recognition in noise: A comparison of children with cochlear implants and children with typical hearing

2024-12-23T10:41:50+01:00

Our objective was to examine how children with bilateral cochlear implants (CIs) recognize the five Spanish vowels in different noise conditions, in comparison to children with typical hearing (TH). The stimuli (the syllables [da], [de], [di], [do], [du]) were embedded in either multi-speaker background babble or speech-shaped noise at the signal-to-noise ratios (SNR) of 0, 6, and 12 dB (forced-choice paradigm). Main effects for Group (of children), Vowel, NoiseType, and SNR were found significant as well as interactions between Group*Vowel and NoiseType*SNR. These results toge-ther with the matrix analyses revealed high error rates for [u] in both groups of children, for [o] in children with CIs only, and frequent confusions of [u] and [o] in both groups, but especially for children with CIs. We conclude that recognizing certain vowels can be challenging in the presence of noise, for children in general but for children with CIs in particular.

Methodology implemented in the collection of a corpus of dialogued speech for the phonosyntactic study of Chilean Spanish

2024-12-15T20:42:36+01:00

This article describes the methodology implemented for the collection of a dialogical corpus of Chilean Spanish, divided into two groups: the first was collected from free conversation and the second from three tests to elicit interaction. For the first corpus, with less control, conversations were initiated based on contingent topics; for the second, three tests were applied to motivate dialogues, these were MapTask (Anderson et al., 1991), Diapix (Van Engen et al., 2010) and the Test of Elicitation of Argumentative Statements (PEEA) adapted from the DELE test, A2 (Martínez et al., 2014). The objective of this corpus is to have samples of connected speech for the study of phonetic and phonological aspects, such as the elements of continuous speech and syllabic restructuring.

Exploring Forensic Phonetics: Understanding Speech Intelligibility in Varied Acoustic Environments

2025-02-07T12:35:49+01:00

In forensic linguistics and more specifically in the field of forensic phonetics, the environment in which a recording is made exerts a considerable influence. This impact arises not only from the acoustic conditions of the space but also from how these conditions affect speech articulation and intelligibility. The way in which sounds are produced, transmitted and perceived can be significantly altered by factors such as background noise, reverberation or the proximity of the speaker to the recording device. This research delves into the analysis of speech intelligibility, as well as vocal characteristics such as f₀ and intensity, to demonstrate the intricate relationship between the concept of intelligibility and various vocal features, showing how these elements are interconnected and how they contribute to clearer or more accurate speech interpretation in forensic contexts.

Duration of stressed vowels in adverbs ending in -mente: data from the PRESEEA-Las Palmas corpus

2025-06-17T21:57:09+02:00

The present study analyses the duration of stressed vowels in a sample of 717 ‑mente adverbs extracted from the PRESEEA-Las Palmas corpus. The findings show that the simultaneous presence of two stresses leads to a reduction in the duration of stressed vowels through a process of temporal readjustment. This process attenuates vowel lengthening, thereby avoiding a doubling of the overall time and promoting greater isochrony within the word. It has also been determined that the stressed vowel of the ending does not invariably exhibit a longer duration than the vowel of the adjective base, as its length is contingent on internal factors. This phenomenon of temporal compensation is evident in the observation that the lengthening of one of the stressed vowel results in a reduction in the temporal space corresponding to the other stressed vowel. These displacements contribute to balancing the overall duration that stress prominence assigns to stressed vowels.

Atlantic Spanish variation: acoustic analysis of consonants across five coastal cities in Hispanic America, integrating sociolinguistic variables

2025-07-13T19:11:21+02:00

This study empirically examines pronunciation patterns in five Hispanic American coastal cities: Camagüey (Cuba), Consuelo (Dominican Republic), Caracas (Venezuela), Buenos Aires (Argentina), and Punta Arenas (Chile). Using data from the Tierras Altas y Bajas de Hispanoamérica (TiAlBa) corpus, it analyzes the realization of /s/ (< s >), /x/ (, before front vowels), and intervocalic /d/ () with respect to linguistic and sociolinguistic factors. The results show that the variables conditioning pronunciation differ across the target consonants. The diatopic variable emerges as the most decisive, with significant effects for all three. In contrast, age and gender do not exert a significant influence. With regard to the linguistic factors –speech style (read word list, read text, guided interview), phonological context, and syllable stress– the effects differ across the consonants, as multiple variables prove significant for each consonant but in different constellations.