Hate speech operationalization: a preliminary examination of hate speech indicators and their structure

Hate speech should be tackled and prosecuted based on how it is operationalized. However, the existing theoretical definitions of hate speech are not sufficiently fleshed out or easily operable. To overcome this inadequacy, and with the help of interdisciplinary experts, we propose an empirical definition of hate speech by providing a list of 10 hate speech indicators and the rationale behind them (the indicators refer to specific, observable, and measurable characteristics that offer a practical definition of hate speech). A preliminary exploratory examination of the structure of hate speech, with the focus on comments related to migrants (one of the most reported grounds of hate speech), revealed that two indicators in particular, denial of human rights and promoting violent behavior, occupy a central role in the network of indicators. Furthermore, we discuss the practical implications of the proposed hate speech indicators—especially (semi-)automatic detection using the latest natural language processing (NLP) and machine learning (ML) methods. Having a set of quantifiable indicators could benefit researchers, human right activists, educators, analysts, and regulators by providing them with a pragmatic approach to hate speech assessment and detection.

VISUAL CUES AND RATER PERCEPTIONS OF SECOND LANGUAGE COMPREHENSIBILITY, ACCENTEDNESS, AND FLUENCY

This study examined the role of visual cues (facial expressions and hand gestures) in second language (L2) speech assessment. University students (N = 60) at English-medium universities assessed 2-minute video clips of 20 L2 English speakers (10 Chinese and 10 Spanish speakers) narrating a personal story. They rated the speakers’ comprehensibility, accentedness, and fluency using 1,000-point sliding scales. To manipulate access to visual cues, the raters were assigned to three conditions that presented audio along with (a) the speaker’s static image, (b) a static image of a speaker’s torso with dynamic face, or (c) dynamic torso and face. Results showed that raters with access to the full video tended to perceive the speaker as more comprehensible and significantly less accented compared to those who had access to less visually informative conditions. The findings are discussed in terms of how the integration of visual cues may impact L2 speech assessment.

Vocabulary Acquisition and Usage for Late Talkers Treatment: Effect on Expressive Vocabulary and Phonology

Purpose This study examined the effect of Vocabulary Acquisition and Usage for Late Talkers (VAULT) treatment on toddlers' expressive vocabulary and phonology. Parent acceptability of VAULT treatment was also considered. Method We used a nonconcurrent multiple baseline single case experimental design with three late talking toddlers aged 21–25 months. The treatment was delivered twice weekly in 30-min sessions for 8 weeks by a rotating team of four speech-language pathologists. Toddlers heard three of their 10 strategically selected target words a minimum of 64 times in play activities each session. Expressive vocabulary and phonology was assessed pre–post, with parent interviews conducted posttreatment. Results All toddlers increased production of target words and expressive vocabulary. Ambient expressive vocabulary size increased by an average of 16 words per week (range of 73–169 words learned over the treatment period). On a 20-item, single-word speech assessment, the toddlers' phonetic inventories increased on average from three to seven consonants, and five to eight vowels. Two toddlers used protowords pretreatment, which were replaced by recognizable attempts at words posttreatment. Parents reported the treatment was acceptable for the child and their family with future consideration of parent-based delivery of the treatment in the home. Conclusions The results of this treatment provide further evidence of a model of intervention informed by the principles of implicit learning, and the interconnectedness of phonological and lexical learning. Investigation is required to establish the efficacy and feasibility of VAULT in clinical contexts. Supplemental Material https://doi.org/10.23641/asha.14714733

Aerosol-generating behaviours in speech pathology clinical practice: A systematic literature review

Objective To evaluate the evidence of aerosol generation across tasks involved in voice and speech assessment and intervention, to inform better management and to reduce transmission risk of such diseases as COVID-19 in healthcare settings and the wider community. Design Systematic literature review. Data sources and eligibility Medline, Embase, Scopus, Web of Science, CINAHL, PubMed Central and grey literature through ProQuest, The Centre for Evidence-Based Medicine, COVID-Evidence and speech pathology national bodies were searched up until August 13th, 2020 for articles examining the aerosol-generating activities in clinical voice and speech assessment and intervention within speech pathology. Results Of the 8288 results found, 39 studies were included for data extraction and analysis. Included articles were classified into one of three categories: research studies, review articles or clinical guidelines. Data extraction followed appropriate protocols depending on the classification of each article (e.g. PRISMA for review articles). Articles were assessed for risk of bias and certainty of evidence using the GRADE system. Six behaviours were identified as aerosol generating. These were classified into three categories: vegetative acts (coughing, breathing), verbal communication activities of daily living (speaking, loud voicing), and performance-based tasks (singing, sustained phonation). Certainty of evidence ranged from very low to moderate with variation in research design and variables. Conclusions This body of literature helped to both identify and categorise the aerosol-generating behaviours involved in speech pathology clinical practice and confirm the low level of evidence throughout the speech pathology literature pertaining to aerosol generation. As many aerosol-generating behaviours are common human behaviours, these findings can be applied across healthcare and community settings. Systematic review registration Registration number CRD42020186902 with PROSPERO International Prospective Register for Systematic Reviews.

Performance Evaluation of Machine Learning Frameworks for Aphasia Assessment

Speech assessment is an essential part of the rehabilitation procedure for patients with aphasia (PWA). It is a comprehensive and time-consuming process that aims to discriminate between healthy individuals and aphasic patients, determine the type of aphasia syndrome, and determine the patients’ impairment severity levels (these are referred to here as aphasia assessment tasks). Hence, the automation of aphasia assessment tasks is essential. In this study, the performance of three automatic speech assessment models based on the speech dataset-type was investigated. Three types of datasets were used: healthy subjects’ dataset, aphasic patients’ dataset, and a combination of healthy and aphasic datasets. Two machine learning (ML)-based frameworks, classical machine learning (CML) and deep neural network (DNN), were considered in the design of the proposed speech assessment models. In this paper, the DNN-based framework was based on a convolutional neural network (CNN). Direct or indirect transformation of these models to achieve the aphasia assessment tasks was investigated. Comparative performance results for each of the speech assessment models showed that quadrature-based high-resolution time-frequency images with a CNN framework outperformed all the CML frameworks over the three dataset-types. The CNN-based framework reported an accuracy of 99.23 ± 0.003% with the healthy individuals’ dataset and 67.78 ± 0.047% with the aphasic patients’ dataset. Moreover, direct or transformed relationships between the proposed speech assessment models and the aphasia assessment tasks are attainable, given a suitable dataset-type, a reasonably sized dataset, and appropriate decision logic in the ML framework.