Speech Content Research Articles

A longitudinal investigation of the acoustic properties of infant-directed speech from 6 to 18 months.

Caregivers often modulate their speech when interacting with infants, adapting a register that has been suggested to have attentional, affective and didactic purposes. The present preregistered study examined the longitudinal trajectories of a diverse range of acoustic features of infant-directed speech (IDS) and compared these with adult-directed speech (ADS), in Norwegian parents of 6- to 18-month-old infants. Sixty-nine families participated. Throughout five laboratory visits across one year, parents were recorded reading a picture-book to their infant (IDS) and an experimenter (ADS). The book was designed to tightly control for the linguistic content and context of speech between participants, timepoints and registers. Analyses of a total of 54 594 vowels and 22 958 phrases revealed, first, an overall effect of register: parents used higher pitch, wider pitch range, slower articulation rate, longer vowel duration and more variable and less distinct vowels in IDS than in ADS. Second, significant register-by-age interactions indicated that parents' IDS, compared with their ADS, featured wider pitch range, larger vowel space and shorter vowel duration in older as compared with younger infants, while pitch, articulation rate and vowel variability and distinctiveness remained relatively stable with age. Results are discussed in the context of the proposed functions of IDS.

The Complexity of Freedom of Speech: Wittgenstein’s Language Game and Gadamer’s Hermeneutics

Freedom of speech is filled in various facets of daily life, which has long been the focus of people’s attention and debate. The concept has multiple layers, including mode, content, and timing of speech. It is worth noting that in different cultural, political and legal contexts, different groups of people interpret and apply freedom of speech in various ways. The purpose of this paper is to compare and analyze Austin’s Speech Act Theory, Ludwig Wittgenstein’s Language Game Theory and Hans-Georg Gadamer’s Hermeneutics. Also it aims to demonstrate and summarize three philosophers’ views on freedom of speech from multiple perspectives. By examining these philosophical frameworks, this paper attempts to guide people to further understand the factors that influence freedom of speech and clarify its inherent complexity. This paper not only provides a detailed analysis of the role of language in different contexts, but also provides a relatively complete model for understanding the dynamic interaction between language and society. In the era of globalization, the contradiction between the basic principle and practice of freedom of expression has become increasingly prominent. The results of the paper are intended to provide a reference for future efforts to strengthen intercultural dialogue to guarantee freedom of speech in pluralistic societies.

Qualitative Analysis of Responses in Estimating Older Adults Cognitive Functioning in Spontaneous Speech: Comparison of Questions Asked by AI Agents and Humans.

Artificial Intelligence (AI) technology is gaining attention for its potential in cognitive function assessment and intervention. AI robots and agents can offer continuous dialogue with the elderly, helping to prevent social isolation and support cognitive health. Speech-based evaluation methods are promising as they reduce the burden on elderly participants. AI agents could replace human questioners, offering efficient and consistent assessments. However, existing research lacks sufficient comparisons of elderly speech content when interacting with AI versus human partners, and detailed analyses of factors like cognitive function levels and dialogue partner effects on speech elements such as proper nouns and fillers. This study investigates how elderly individuals' cognitive functions influence their communication patterns with both human and AI conversational partners. A total of 34 older people (12 men and 22 women) living in the community were selected from a silver human resource centre and day service centre in Tokyo. Cognitive function was assessed using the Mini-Mental State Examination (MMSE), and participants engaged in semi-structured daily conversations with both human and AI partners. The study examined the frequency of fillers, proper nouns, and "listen back" in conversations with AI and humans. Results showed that participants used more fillers in human conversations, especially those with lower cognitive function. In contrast, proper nouns were used more in AI conversations, particularly by those with higher cognitive function. Participants also asked for explanations more often in AI conversations, especially those with lower cognitive function. These findings highlight differences in conversation patterns based on cognitive function and the conversation partner being either AI or human. These results suggest that there are differences in conversation patterns depending on the cognitive function of the participants and whether the conversation partner is a human or an AI. This study aims to provide new insights into the effective use of AI agents in dialogue with the elderly, contributing to the improvement of elderly welfare.

“Ask” or “Inquire”: operationalizing speech formality in psychosis and its risk states using etymology

Many individuals with psychotic symptoms have less complex language than healthy individuals. Word etymology is a lexical feature that has not yet been studied in clinical populations, but among healthy individuals, words of Old French origin are chosen over Germanic-origin words to convey formality (e.g. “inquire” vs. “ask”). Differences in language complexity among individuals with psychotic symptoms may relate to differences in etymological content in speech. Here, we determined the proportion of Germanic-origin word use and Old-French-origin word use in a large cohort of individuals with recent-onset psychosis or at clinically high risk for psychosis, hypothesizing that individuals with recent onset psychosis would have increased use of Germanic-origin words and decreased use of Old-French-origin words. This hypothesis was borne out, even after adjusting for sex, age, recruitment site, education, racial identity, and for a subset, IQ. Etymology proportions were associated with role but not social functioning in individuals with psychotic symptoms, consistent with the premise that they reflect speech formality. Understanding speech differences in the psychosis spectrum through the lens of etymology may lead to new interventions to improve role functioning.

Under-Represented Speech Dataset from Open Data: Case Study on the Romanian Language

This paper introduces the USPDATRO dataset. This is a speech dataset, in the Romanian language, constructed from open data, focusing on under-represented voice types (children, young and old people, and female voices). The paper covers the methodology behind the dataset construction, specific details regarding the dataset, and evaluation of existing Romanian Automatic Speech Recognition (ASR) systems, with different architectures. Results indicate that more under-represented speech content is needed in the training of ASR systems. Our approach can be extended to other low-resourced languages, as long as open data are available.

Dual-stream Noise and Speech Information Perception based Speech Enhancement

In real-world scenarios, dynamic ambient noise often degrades speech quality, highlighting the need for advanced speech enhancement techniques. Traditional methods, which rely on static embeddings as auxiliary features, struggle to address the complexities of varying noise conditions. To overcome this, we propose a Dual-stream Noise and Speech Information Perception (DNSIP) approach that dynamically detects and processes both noise and speech through innovative information extraction and suppression mechanisms. Initially, non-speech segments predominantly contain environmental noise, while speech segments carry information about the intended speaker. To handle this dynamic nature, real-time voice activity detection (VAD) is employed to accurately differentiate between speech and noise components. Building on VAD estimates, we propose an innovative information extraction framework that selectively extracts relevant noise and speech features from the noisy input, establishing a dual-stream network for concurrent noise and speech learning. To account for the temporal and spectral variability of noise and speech, a frequency-sequence attention mechanism is integrated, enhancing the model’s ability to learn contextual and spectral dependencies. Additionally, an information suppression module is introduced to minimize cross-stream interference by attenuating noise within the speech stream and suppressing speech content within the noise stream. The derived noise and speech spectrograms are then utilized to formulate a minimum mean square error log-spectral amplitude (MMSE-LSA) estimator for robust speech enhancement. Experimental evaluations on the WSJ0 and VCTK+DEMAND datasets demonstrate that our DNSIP approach surpasses existing state-of-the-art methods, underscoring its efficacy in challenging acoustic environments.

Dog-human vocal interactions match dogs' sensory-motor tuning.

Within species, vocal and auditory systems presumably coevolved to converge on a critical temporal acoustic structure that can be best produced and perceived. While dogs cannot produce articulated sounds, they respond to speech, raising the question as to whether this heterospecific receptive ability could be shaped by exposure to speech or remains bounded by their own sensorimotor capacity. Using acoustic analyses of dog vocalisations, we show that their main production rhythm is slower than the dominant (syllabic) speech rate, and that human-dog-directed speech falls halfway in between. Comparative exploration of neural (electroencephalography) and behavioural responses to speech reveals that comprehension in dogs relies on a slower speech rhythm tracking (delta) than humans' (theta), even though dogs are equally sensitive to speech content and prosody. Thus, the dog audio-motor tuning differs from humans', and we hypothesise that humans may adjust their speech rate to this shared temporal channel as means to improve communication efficacy.

Decoding Silent Speech Cues From Muscular Biopotential Signals for Efficient Human‐Robot Collaborations

AbstractSilent speech interfaces offer an alternative and efficient communication modality for individuals with voice disorders and when the vocalized speech communication is compromised by noisy environments. Despite the recent progress in developing silent speech interfaces, these systems face several challenges that prevent their wide acceptance, such as bulkiness, obtrusiveness, and immobility. Herein, the material optimization, structural design, deep learning algorithm, and system integration of mechanically and visually unobtrusive silent speech interfaces are presented that can realize both speaker identification and speech content identification. Conformal, transparent, and self‐adhesive electromyography electrode arrays are designed for capturing speech‐relevant muscle activities. Temporal convolutional networks are employed for recognizing speakers and converting sensing signals into spoken content. The resulting silent speech interfaces achieve a 97.5% speaker classification accuracy and 91.5% keyword classification accuracy using four electrodes. The speech interface is further integrated with an optical hand‐tracking system and a robotic manipulator for human‐robot collaboration in both assembly and disassembly processes. The integrated system achieves the control of the robot manipulator by silent speech and facilitates the hand‐over process by hand motion trajectory detection. The developed framework enables natural robot control in noisy environments and lays the ground for collaborative human‐robot tasks involving multiple human operators.

Combining low-technology augmentative and alternative communication with regular aphasia treatment: interim findings of a hospital-based RCT in post-stroke patients

ABSTRACT Background Previous research indicates that low-technology (low-tech) augmentative and alternative communication interventions (AACT) can enhance communication skills in individuals with post-stroke aphasia in China. However, there is limited evidence for low-tech AACT in inpatient settings. Aim This study aims to evaluate the impact of low-tech AACT combined with speech-language therapy (SLT) on communication skills related to basic needs, quality of life, and spoken language function among individuals with post-stroke aphasia in an inpatient setting. Methods & Procedures Twelve participants with aphasia having difficulty in daily communication were randomly assigned to an experimental group, receiving 30 minutes of SLT and 30 minutes of low-tech AACT per session (AACT + SLT), or a control group, receiving 60 minutes of SLT per session (SLT), over ten sessions in two weeks. Primary outcomes were measured using the basic needs (CBN) domain of the functional assessment of communication skills for adults (FACS). Secondary outcomes included scores from the Chinese version of the stroke-specific quality of life scale (SS-QOL) and subtests of spoken language production and comprehension from the aphasia battery of Chinese (ABC). The study was registered in the Chinese clinical trial registry (ChiCTR2000028870). Results The results indicated a significant improvement in the CBN domain of FACS in the AACT + SLT group, with larger clinically significant changes compared to the SLT group. The AACT+ SLT group also showed significantly higher overall scores on the SS-QOL and more significant improvements in the content of spontaneous speech than the SLT group after intervention. Comparable improvements were observed in other subtests of the ABC between the two groups. Conclusions Incorporating low-tech AACT into regular SLT in inpatient settings can improve communication for basic needs in an in-patient setting, facilitate language recovery, and potentially support the overall quality of life, compared to SLT alone, for people with post-stroke aphasia. This combined approach demonstrates preliminary feasibility in hospital settings, suggesting the value of low-tech AACT in aphasia rehabilitation.

Zero-shot voice conversion based on feature disentanglement

Voice conversion (VC) aims to convert the voice from a source speaker to a target speaker without modifying the linguistic content. Zero-shot voice conversion has attracted significant attention in the task of VC because it can achieve conversion for speakers who did not appear during the training stage. Despite the significant progress made by previous methods in zero-shot VC, there is still room for improvement in separating speaker information and content information. In this paper, we propose a zero-shot VC method based on feature disentanglement. The proposed model uses a speaker encoder for extracting speaker embeddings, introduces mixed speaker layer normalization to eliminate residual speaker information in content encoding, and employs adaptive attention weight normalization for conversion. Furthermore, dynamic convolution is introduced to improve speech content modeling while requiring a small number of parameters. The experiments demonstrate that performance of the proposed model is superior to several state-of-the-art models, achieving both high similarity with the target speaker and intelligibility. In addition, the decoding speed of our model is much higher than the existing state-of-the-art models.

Human detection of political speech deepfakes across transcripts, audio, and video

Recent advances in technology for hyper-realistic visual and audio effects provoke the concern that deepfake videos of political speeches will soon be indistinguishable from authentic video. We conduct 5 pre-registered randomized experiments with N = 2215 participants to evaluate how accurately humans distinguish real political speeches from fabrications across base rates of misinformation, audio sources, question framings with and without priming, and media modalities. We do not find base rates of misinformation have statistically significant effects on discernment. We find deepfakes with audio produced by the state-of-the-art text-to-speech algorithms are harder to discern than the same deepfakes with voice actor audio. Moreover across all experiments and question framings, we find audio and visual information enables more accurate discernment than text alone: human discernment relies more on how something is said, the audio-visual cues, than what is said, the speech content.

Engagement of the speech motor system in challenging speech perception: Activation likelihood estimation meta-analyses.

The relationship between speech production and perception is a topic of ongoing debate. Some argue that there is little interaction between the two, while others claim they share representations and processes. One perspective suggests increased recruitment of the speech motor system in demanding listening situations to facilitate perception. However, uncertainties persist regarding the specific regions involved and the listening conditions influencing its engagement. This study used activation likelihood estimation in coordinate-based meta-analyses to investigate the neural overlap between speech production and three speech perception conditions: speech-in-noise, spectrally degraded speech and linguistically complex speech. Neural overlap was observed in the left frontal, insular and temporal regions. Key nodes included the left frontal operculum (FOC), left posterior lateral part of the inferior frontal gyrus (IFG), left planum temporale (PT), and left pre-supplementary motor area (pre-SMA). The left IFG activation was consistently observed during linguistic processing, suggesting sensitivity to the linguistic content of speech. In comparison, the left pre-SMA activation was observed when processing degraded and noisy signals, indicating sensitivity to signal quality. Activations of the left PT and FOC activation were noted in all conditions, with the posterior FOC area overlapping in all conditions. Our meta-analysis reveals context-independent (FOC, PT) and context-dependent (pre-SMA, posterior lateral IFG) regions within the speech motor system during challenging speech perception. These regions could contribute to sensorimotor integration and executive cognitive control for perception and production.

FOLK THINKING AS A PSYCHOLOGICAL TOOL FOR THE REGULATION OF ETHNIC BEHAVIOR

The article refers to the verbal formulaic structures that strengthen the self-composition of a person in extreme situations, which have been crystalized and perceived by the people over time, becoming a unique tool for self-defense. Speech practices are very important from the point of view of the preservation of the given ethnos and as embodiments of clarification of thought, they are reflected in the decisive moments in the life of the nation or people, especially in the context of great psychological tension (natural disaster, flood, disaster, war, etc.). We see the formation of a verbal saying as a unique result of the perceptive experience of the eternal mind or as a type of oral tradition that arose on the basis of archetypal thinking. The phrases in question have accompanied the humankind through history and have been conceived by practices over the centuries, being accepted as traditional behaviors in the given community, they influence the behavior of ethnocrats and provide super-orienting references. Stable phrases awaken stronger voice than our own: the speaker with folk oral thinking speaks with a thousands of voices, captivates and humbles, elevates what is described from the singular and temporary to the realm of eternal existence. In an extreme state, we implement the forms and content of speech that our mental world needs and that correspond to the content of our life.

EarSpeech: Exploring In-Ear Occlusion Effect on Earphones for Data-efficient Airborne Speech Enhancement

Earphones have become a popular voice input and interaction device. However, airborne speech is susceptible to ambient noise, making it necessary to improve the quality and intelligibility of speech on earphones in noisy conditions. As the dual-microphone structure (i.e., outer and in-ear microphones) has been widely adopted in earphones (especially ANC earphones), we design EarSpeech which exploits in-ear acoustic sensory as the complementary modality to enable airborne speech enhancement. The key idea of EarSpeech is that in-ear speech is less sensitive to ambient noise and exhibits a correlation with airborne speech. However, due to the occlusion effect, in-ear speech has limited bandwidth, making it challenging to directly correlate with full-band airborne speech. Therefore, we exploit the occlusion effect to carry out theoretical modeling and quantitative analysis of this cross-channel correlation and study how to leverage such cross-channel correlation for speech enhancement. Specifically, we design a series of methodologies including data augmentation, deep learning-based fusion, and noise mixture scheme, to improve the generalization, effectiveness, and robustness of EarSpeech, respectively. Lastly, we conduct real-world experiments to evaluate the performance of our system. Specifically, EarSpeech achieves an average improvement ratio of 27.23% and 13.92% in terms of PESQ and STOI, respectively, and significantly improves SI-SDR by 8.91 dB. Benefiting from data augmentation, EarSpeech can achieve comparable performance with a small-scale dataset that is 40 times less than the original dataset. In addition, we validate the generalization of different users, speech content, and language types, respectively, as well as robustness in the real world via comprehensive experiments. The audio demo of EarSpeech is available on https://github.com/EarSpeech/earspeech.github.io/.

Code-Mixing Used by Alexandra Asmasoebrata in Volix Media Youtube Channel

Sociolinguistics is the study that concerns the relation between society and language use. The part of sociolinguistics that is code-mixing The study aims at (1) identifying the types, and (2) defining the factors of code mixing used by Alexandra Asmasoebrata on the Volix Media YouTube Channel. The focused of the study is on code mixing Indonesian-English. The data was collected by Volix Media YouTube channel. This study employs a qualitative research methodology, and the observation method is the data collection strategy. Hoffman's (1991) theory is used to analyze the many forms and factors of code mixing. This study shows that there are 11 utterances were used by Alexandra Amasoebrata on the Volix Media Youtube channel that contained code-mixing. 4 data for intra-sentential, 2 data for intra-lexical. The factor found in the data source is 2 data for talking about a particular topic, 1 data for repetition used for clarification, the intention of clarifying the speech content for interlocutor is 1 data, and being emphatic about something (expressing solidarity) is 1 data. Keywords: Sociolinguistic, Code-mixing, Youtube

Code Mixing Used by Yuki Kato in Bicara Cinta on Puella Id Youtube Channel

Code-mixing gives an understanding of a mix of two languages or more by the multilingual. This study aims at defining the types and the factors causing the code-mixing used by Yuki Kato in a Youtube video. A mixed method was applied, with both qualitative and quantitative methods used in the study. The data is all of the words, phrases, or sentences that include the use of mixed codes in English and Indonesian Language, which was collected from one episode of Bicara Cinta on the PUELLA ID Youtube Channel. The code-mixing theory from Hoffman (1991) was used to analyze the data, including the types of code-mixing and the factors that caused the code-mixing. The analysis has identified 93 data on the use of code-mixing by Yuki Kato, including three types of code-mixing which are intra-sentential with 80 data found (86,02%), intra-lexical with 10 data found (10, 75%), and involving a change of pronunciation with 3 data found (3,23%). Six out of seven factors were applied in the code-mixing by Yuki Kato. The code-mixing produced by Yuki Kato is caused by several factors, those are talking about particular topic with 46 data found (49,46%), being emphatic about something with 35 data found (37,63%), intention of clarifying the speech content for the interlocutor with 7 data found (7,53%), repetition used for clarification with 2 data found (2,15%), interjection with 2 data found (2,15%), and quoting somebody else with only 1 data found (1,08%). There is no code-mixing caused by expressing a group identity. Keywords: Code-mixing, Codes, Youtube Channel

Code-Switching and Code-Mixing in Sesotho Accordion Music: A Sociolinguistic Perspective

This article explores the use of code-switching/mixing in Sesotho accordion music within a sociolinguistic framework. Code-switching/mixing is one of the sociolinguistic concepts that is mainly highlighted in conversation. It refers to switching or mixing between two or more languages in the context of a single conversation or situation. Just like a conversation, Sesotho accordion music also exhibits the phenomenon of code-switching/mixing. The Basotho accordion music (’mino oa koriana) is one of the forms of the oral tradition from Lesotho whose singing is accompanied by a homemade drum, and it has its own unique and acceptable taste. This music is seen as both captivating and complex. We observe that its fascination and its complex features are a result of the incorporation of code-switching/mixing in a single song. We know how the songs are generated, and they are part of the generic culture and popular culture of the people. We purposively selected four songs from recorded tapes that used code-switching/mixing. Using critical discourse analysis, the results reveal that code-switching/mixing in the accordion music of the Basotho reflects intertextuality, which includes borrowing, inclusion and exclusion, literary and conversational intertextuality. The study also shows that code-switching/mixing reflects repetition used for emphasis, clarity or amplification. The findings further express affirmation of class/age identity and the intention of clarifying the speech content for the interlocutor. One of the emerging issues is that change in pronunciation may reflect tone and register in the accordion music of the Basotho.

The Impact of Code Switching and Code Mixing on Learning of English Language in Lower Secondary School level: A Case Study of Selected Schools in Rubavu District, Rwanda

English is the most widely spoken language in the world, making it essential for effective global communication. This is why individuals who aspire to thrive in today's global society must possess proficiency in English. This study focuses on how the use of code switching (CS) and code mixing (CM) affect the learning of the English language in G. S. Bisizi and G. S. Gikombe of Rubavu District in Rwanda. The study was conducted with the aim to: (1) identify reasons for code switching and code mixing in lower secondary schools; (2) assess the impact of CS and CM on students’ performance in learning English; and (3) establish strategies to adopt in order to avoid CS and CM in English language classrooms. It was conducted based on the matrix language frame and sequential theories. A descriptive research design was used. The researchers employed a mixed method for data collection, analysis, and presentation. Its total population is comprised of 726 individuals, including 716 students, 6 teachers, and 4 school leaders from both schools. The sample size is 85 respondents (75 students, 6 teachers, and 4 school leaders) got by use of simple random sampling (students) and purposive sampling techniques (teachers and school leaders). During the collection of primary data, questionnaires, interview guides, and observation checklists were used as convenient tools, while documentation was used in collecting secondary data. The findings indicated that the lack of vocabulary, emphasizing a particular topic while teaching and learning English, and the intention to clarify the speech content are the major reasons for using CS and CM. They also reveal that CS and CM contribute to students’ failure in learning English, as both create a lack of their confidence in speaking, limit students’ practice in speaking English, retard their ability to master English, and lead to a low understanding and mastery of English. In addition, the study revealed that teachers are the main source for CS and CM in their classes. Finally, this research recommends that school leaders and English language teachers should reinstate a policy of speaking English in the school premises and make an effective follow-up to ensure students do not violate it. Teachers should also act as role models in their language classes by avoiding CS and CM, whereas the decision makers in education should initiate a short-term professional program and special ways to avoid CS among teachers to help students effectively learn English language.

Enhancing speech emotion detection with Windowed Long-Term Average Spectrum and Logistic-Rectified Linear Unit

Speech emotion detection can help monitor people's mental health by analyzing their speech patterns and recognizing emotional indicators. The raw data is initially augmented using various techniques such as shifting, pitching, noise injection, and stretching. Some feature extraction methods may only produce a limited set of features, potentially missing out on important information present in the data. This can result in reduced accuracy and the inability to capture complex patterns and variations in speech signals. The planned work consists of two phases: attribute extraction and classification. In the first phase, the WLTAS (Windowed Long-Term Average Spectrum) feature extraction extracts spectral information from speech signals, providing a compact and representative feature representation. It captures the frequency content and energy distribution of speech over time, enabling the analysis of important spectral characteristics. In the second phase, the Logistic-Rectified Linear Unit (LoRLU) introduces sparsity. It enhances the discriminative power of the classifier by allowing only positive values to pass through while discarding negative values. The Gibbs Restricted Boltzmann Machine (GRBM) is a generative probabilistic model that studies a joint probability delivery of the input data. It is particularly effective in capturing dependencies and patterns in high-dimensional data. The performance analysis of WLTAS-L-GRBM (Windowed Long-Term Average Spectrum - Logistic-Rectified Linear Unit - Gibbs Restricted Boltzmann Machine) has been conducted using evaluation metrics such as accuracy (99%), precision (0.97), recall (1), F1 score (0.98), MSE (0.01), and AUC-ROC (0.97) and also evaluated against various speech emotion datasets.

Appraisal Theory and Interpreting Political Speech

Appraisal theory can be used to study the semiotic content of formal speeches. This paper focuses on the study of two speeches by Ban Ki Moon. The first in 2006 was his acceptance speech and the second, in 2016, was his speech on leaving his role as Secretary General of the United Nations. By employing the same speaker and organization, this helps remove variances due to personal or corporate style. It also allows a focus on how different semiotic approaches are used at different stages of each speech. While affect is common across the speech structure, judgement is more commonly found in sections that can be seen as instances of expected politeness towards the intended audience. By using such modes, the speaker is effectively inviting agreement with their overall presentation. This also suggests a linkage between chosen mode and purpose. However, meaning-making cannot just be understood through a sentence by sentence analysis and there still remains a need to take a holistic overview and place the speech in a wider context.