This survey study examines speech-language pathologists' (SLPs) experiences with teletherapy service delivery at the onset of the COVID-19 pandemic. It also explores the potential for integrating teletherapy into school-based therapeutic services beyond pandemic-related circumstances. A 56-question digital survey was administered between April 20 and June 4, 2020, to members of select American Speech-Language-Hearing Association (ASHA) community boards. SLPs reported on their pandemic preparation to initiate teletherapy, clients' eligibility for teletherapy, clients' technological readiness for teletherapy, perceptions of communication differences with parents/caregivers, and perceptions of parent/caregiver engagement during in-person and teletherapy service delivery. A total of 470/655 responses were analyzed. Most SLPs prepared for teletherapy through webinars, ASHA website resources, and collaboration with colleagues. Key factors influencing client candidacy for teletherapy included behaviors such as uncooperativeness (65.0%); inattentiveness (55.6%); distractibility (44.4%); and speech-language conditions, specifically pragmatic functions (42.0%). Many SLPs (75.2%) noted that families lacked adequate access to telecommunication resources for remote therapy. Despite this, teletherapy improved communication with parents regarding students' therapy progress and carryover strategies. SLPs also reported feeling more comfortable with parent/caregiver participation in teletherapy as compared to in-person sessions. However, challenges were noted, including difficulties with setting up equipment and communicating with non–English-speaking families. Teletherapy provided significant benefits to SLPs and their school-aged clients during the pandemic. However, integrating teletherapy into school-based settings requires careful planning and consideration. We recommend further empirical research to evaluate its effectiveness across different school levels and speech-language diagnostic conditions in school-aged children. https://doi.org/10.23641/asha.29100743
Subsequent to the COVID-19 pandemic, face masks remain a consistent part of many individuals’ lives. While research on face masking has revealed effects on speech intelligibility, relatively few studies have examined speaker differences, the effects of masking on speaker behavior, and how these influence intelligibility. We report here on data from four healthy American English-speaking adults who produced low-predictability sentences in four face mask conditions: surgical, cloth, N95, and no mask. Typically hearing listeners (n = 89) heard sentences, mixed with multi-talker babble, from one of the speakers and typed the words they heard. Percent words correct was scored by a text-matching algorithm, allowing for morphological errors and homonyms. From all speakers, we obtained: long-term average spectra; first and second formant ranges in words with /i, ɑ, o/; sentence durations and average intensities; and phonetic features of common misperceptions. Error rates were highest for the N95 mask. Speakers demonstrated individual patterns of the above acoustic variables. The speaker with the lowest intelligibility displayed faster speech, lower intensity, and unexpectedly higher F2 ranges in the masked conditions. These results suggest that quantifying speaker differences in intelligibility and acoustics can better elucidate the effects of face masks on communication.
Traditionally, metamorphic testing is an approach for quality assurance engineers to generate test cases with oracles for purposes of assuring the functional correctness of a system under test. We provide an argument for utilizing the metamorphic testing technique for the purpose of domain and application specific data synthesis for machine-learning systems, and we propose a process for creating and maintaining a library of metamorphic relations that concisely captures subject matter expert knowledge about classification boundaries in a way that is easy for them to understand and manage. As is typical in machine-learning applications, synthesized ground truth data of the sort can be useful both for training and testing. We illustrate and evaluate the method in an application to detect stealthy host-initiated denial of service or jamming, showing the method can be used to improve attack detectors through training and identify weaknesses through validation.
BACKGROUND: Primary prevention is a desirable path to decrease the incidence of primary cam morphology (PCM). This commentary provides specific attention to peak height velocity timing and training load management during critical growth periods in adolescent athletes. CLINICAL QUESTION: Can physical literacy and load management distribute epiphyseal loads and reduce the occurrence of primary cam morphology through periodized cycles of hip extension? KEY RESULTS AND CLINICAL APPLICATION: We suggest this as a line of research that can be investigated by multiple independent researchers. JOSPT Open 2025;3(4):386-394. Epub 26 May 2025. doi:10.2519/josptopen.2025.0128
Software vulnerabilities are a challenge in cybersecurity. Manual security patches are often difficult and slow to be deployed, while new vulnerabilities are created. Binary code vulnerability detection is less studied and more complex compared to source code, and this has important practical implications. Deep learning has become an efficient and powerful tool in the security domain, where it provides end-to-end and accurate prediction. Modern deep learning approaches learn the program semantics through sequence and graph neural networks, using various intermediate representation of programs, such as abstract syntax trees (AST) or control flow graphs (CFG). Due to the complex nature of program execution, the output of an execution depends on the many program states and inputs. Also, a CFG generated from static analysis can be an overestimation of the true program flow. Moreover, the size of programs often does not allow a graph neural network with fixed layers to aggregate global information. To address these issues, we propose DeepEXE, an agent-based implicit neural network that mimics the execution path of a program. We use reinforcement learning to enhance the branching decision at every program state transition and create a dynamic environment to learn the dependency between a vulnerability and certain program states. An implicitly defined neural network enables nearly infinite state transitions until convergence, which captures the structural information at a higher level. The experiments are conducted on two semi-synthetic and two real-world datasets. We show that DeepEXE is an accurate and efficient method and outperforms the state-of-the-art vulnerability detection methods.
