Advances in Completely Automated Vowel Analysis for Sociophonetics: Using End-to-End Speech Recognition Systems With DARLA.

In recent decades, computational approaches to sociophonetic vowel analysis have been steadily increasing, and sociolinguists now frequently use semi-automated systems for phonetic alignment and vowel formant extraction, including FAVE (Forced Alignment and Vowel Extraction, Rosenfelder et al., 2011; Evanini et al., Proceedings of Interspeech, 2009), Penn Aligner (Yuan and Liberman, J. Acoust. Soc. America, 2008, 123, 3878), and DARLA (Dartmouth Linguistic Automation), (Reddy and Stanford, DARLA Dartmouth Linguistic Automation: Online Tools for Linguistic Research, 2015a). Yet these systems still have a major bottleneck: manual transcription. For most modern sociolinguistic vowel alignment and formant extraction, researchers must first create manual transcriptions. This human step is painstaking, time-consuming, and resource intensive. If this manual step could be replaced with completely automated methods, sociolinguists could potentially tap into vast datasets that have previously been unexplored, including legacy recordings that are underutilized due to lack of transcriptions. Moreover, if sociolinguists could quickly and accurately extract phonetic information from the millions of hours of new audio content posted on the Internet every day, a virtual ocean of speech from newly created podcasts, videos, live-streams, and other audio content would now inform research. How close are the current technological tools to achieving such groundbreaking changes for sociolinguistics? Prior work (Reddy et al., Proceedings of the North American Association for Computational Linguistics 2015 Conference, 2015b, 71-75) showed that an HMM-based Automated Speech Recognition system, trained with CMU Sphinx (Lamere et al., 2003), was accurate enough for DARLA to uncover evidence of the US Southern Vowel Shift without any human transcription. Even so, because that automatic speech recognition (ASR) system relied on a small training set, it produced numerous transcription errors. Six years have passed since that study, and since that time numerous end-to-end automatic speech recognition (ASR) algorithms have shown considerable improvement in transcription quality. One example of such a system is the RNN/CTC-based DeepSpeech from Mozilla (Hannun et al., 2014). (RNN stands for recurrent neural networks, the learning mechanism for DeepSpeech. CTC stands for connectionist temporal classification, the mechanism to merge phones into words). The present paper combines DeepSpeech with DARLA to push the technological envelope and determine how well contemporary ASR systems can perform in completely automated vowel analyses with sociolinguistic goals. Specifically, we used these techniques on audio recordings from 352 North American English speakers in the International Dialects of English Archive (IDEA), extracting 88,500 tokens of vowels in stressed position from spontaneous, free speech passages. With this large dataset we conducted acoustic sociophonetic analyses of the Southern Vowel Shift and the Northern Cities Chain Shift in the North American IDEA speakers. We compared the results using three different sources of transcriptions: 1) IDEA's manual transcriptions as the baseline "ground truth", 2) the ASR built on CMU Sphinx used by Reddy et al. (Proceedings of the North American Association for Computational Linguistics 2015 Conference, 2015b, 71-75), and 3) the latest publicly available Mozilla DeepSpeech system. We input these three different transcriptions to DARLA, which automatically aligned and extracted the vowel formants from the 352 IDEA speakers. Our quantitative results show that newer ASR systems like DeepSpeech show considerable promise for sociolinguistic applications like DARLA. We found that DeepSpeech's automated transcriptions had significantly fewer character error rates than those from the prior Sphinx system (from 46 to 35%). When we performed the sociolinguistic analysis of the extracted vowel formants from DARLA, we found that the automated transcriptions from DeepSpeech matched the results from the ground truth for the Southern Vowel Shift (SVS): five vowels showed a shift in both transcriptions, and two vowels didn't show a shift in either transcription. The Northern Cities Shift (NCS) was more difficult to detect, but ground truth and DeepSpeech matched for four vowels: One of the vowels showed a clear shift, and three showed no shift in either transcription. Our study therefore shows how technology has made progress toward greater automation in vowel sociophonetics, while also showing what remains to be done. Our statistical modeling provides a quantified view of both the abilities and the limitations of a completely "hands-free" analysis of vowel shifts in a large dataset. Naturally, when comparing a completely automated system against a semi-automated system involving human manual work, there will always be a tradeoff between accuracy on the one hand versus speed and replicability on the other hand [Kendall and Joseph, Towards best practices in sociophonetics (with Marianna DiPaolo), 2014]. The amount of "noise" that can be tolerated for a given study will depend on the particular research goals and researchers' preferences. Nonetheless, our study shows that, for certain large-scale applications and research goals, a completely automated approach using publicly available ASR can produce meaningful sociolinguistic results across large datasets, and these results can be generated quickly, efficiently, and with full replicability.

A clinically derived guinea pig dosing model of cisplatin ototoxicity.

The guinea pig is a commonly-used animal model in hearing research, as their audible frequency range is similar to that of humans, and they possess comparatively large cochleae among rodents. Numerous studies have investigated the ototoxic effects of cisplatin in guinea pigs, but these have been mostly limited to single high-dose bolus injections of cisplatin. This method of drug administration is not consistent with human treatment schedules, and therefore lacks translational value to clinical applications. We tested several different cisplatin dosing schedules in guinea pigs based on common research based and clinical regimens, measuring the resulting hearing loss and morbidity (weight loss). We propose a dosing paradigm of once-weekly 4 mg/kg cisplatin injections for three weeks to best mimic clinical treatment schedules. This method resulted in a configuration of hearing loss similar to what is observed in humans along with minimal changes in weight.

Cool OtOprotective Ear Lumen (COOL) Therapy for Cisplatin-induced Hearing Loss.

HYPOTHESIS: Localized cooling of the external ear has a protective effect on the susceptibility to cisplatin-induced hearing loss. BACKGROUND: We previously demonstrated significant protection from cisplatin-induced hearing loss using cool water ear canal irrigation. However, the study was limited to a single bolus injection of cisplatin and an acute time period. Here, we examined the application of localized cooling of the ear canal with repeated doses of cisplatin, over an expanded period of time, and using two methods of cooling. METHODS: Twenty-four guinea pigs (12 male and 12 female) underwent auditory physiological testing (auditory brainstem response and distortion product otoacoustic emissions at 8-32 kHz) and pre/postadministration of cisplatin. Cisplatin (4 mg/kg i.p.) was administered in 3 weekly single injections for a total of 12 mg/kg. While anesthetized, the left ears of the guinea pigs were exposed to either cool water (22°C; ICS Water Caloric Irrigator), a cool ear bar (15°C, cooled by a Peltier device; TNM, Scion NeuroStim), or left uncooled as a sham control. The animals were tested 3 days post each dosage and 1 month post the final dose. At the end of the experiment the animals were euthanized for histological evaluation. RESULTS: We found that hearing loss was significantly reduced, and hair cell survival greatly improved, in animals that received cooling treatments compared to cisplatin-only control animals. No significant difference was observed between the two methods of cooling. CONCLUSION: Localized cooling of the ear canal during administration of cisplatin mitigated loss of auditory function and loss of hair cells.

Can contaminated water be rendered safe for nasal saline irrigations?

OBJECTIVES/HYPOTHESIS: To compare sterile water to three methods of sterilization (carbon filtration, boiling, and ultraviolet [UV] light) for preparation of nasal saline irrigants free of bacterial and amebic contaminants. STUDY DESIGN: Bench-top translational research and cost comparison. METHODS: Sterile water was compared to common sterilization methods. Sterile water was contaminated with known concentrations of Staphylococcus aureus, Pseudomonas aeruginosa, Moraxella catarrhalis, Acinetobacter baumannii, Klebsiella pneumonia, Legionella pneumophila, and Naegleria fowleri. Test samples were subjected to boiling, carbon filtration, or ultraviolet light (UV) and then cultivated on appropriate media. Controls included samples of sterile water (negative control) and untreated test samples (positive control). RESULTS: Carbon filtration reduced but did not eliminate the number of organisms present in test samples. Boiling test samples for 5 minutes and UV light treatment resulted in sterilization of all organisms. Negative (sham contaminated) samples produced no growth, whereas positive (untreated) samples grew numerous organisms as expected. A cost comparison between bottled water and UV water sterilization (with SteriPEN Ultra) became equal in less than 2 years of consistent use. CONCLUSIONS: Carbon filtration reduces contamination but does not sterilize water and is thus unsafe for preparation of nasal saline irrigant. Boiling and UV treatment resulted in sterilization and are equivalent to purchased sterile water. Ultraviolet treatment was found to be safe, convenient, and a cost-effective alternative to purchased sterile water. LEVEL OF EVIDENCE: NA Laryngoscope, 127:1513-1519, 2017.

Case presentation and images of a lingual osseous choristoma in a pediatric patient.

Since its original description in 1913, fewer than 100 lingual osseous choristomas have been reported in the literature; thus, prevalence is unknown. We describe a case of an 11 year old male who was seen in consultation after an incidental left posterior tongue mass was discovered on exam. The patient's presentation of an asymptomatic, hard, pedunculate posterior tongue lesion is typical; however, if one is to believe the proposed congenital remnant theory in regards to the etiology of this benign tumor, it is curious that no mention was made of a lesion of the tongue on prior evaluations by his pediatrician or on the otolaryngologic examinations performed 3 and 6 years prior to the most recent presentation. Included with the case description are interesting radiographs, intra-operative photos, gross specimen photo and microscopic images.

Sex differences in blood pressure control in SHR: lack of a role for EETs.

The mechanisms responsible for the gender difference in blood pressure (BP) in humans are not clear. Over the past several years we have studied the spontaneously hypertensive rat (SHR) as a model of sex differences in BP control. In the present study, we tested the hypothesis that renal vascular and microsomal epoxyeicosatrienoic acid (EET) levels are higher in females than males, and increasing vascular EETs by blocking epoxide hydrolase with AUDA will reduce BP more in males than females. Renal vascular and microsomal EETs were higher in female SHR than males. Mean arterial pressure (MAP by telemetry) was higher in males than females during the baseline period of 6 days, and although the epoxide hydrolase inhibitor, AUDA, given for 10 days increased renal microvascular EETs in both groups, AUDA did not affect MAP in either group. These data suggest that EETs do not contribute to the sex differences in hypertension in young SHR.

A randomized controlled trial examining the efficacy of motivational counseling with observed therapy for antiretroviral therapy adherence.

This study determined whether motivational interviewing-based cognitive behavioral therapy (MI-CBT) adherence counseling combined with modified directly observed therapy (MI-CBT/mDOT) is more effective than MI-CBT counseling alone or standard care (SC) in increasing adherence over time. A three-armed randomized controlled 48-week trial with continuous electronic drug monitored adherence was conducted by randomly assigning 204 HIV-positive participants to either 10 sessions of MI-CBT counseling with mDOT for 24 weeks, 10 sessions of MI-CBT counseling alone, or SC. Poisson mixed effects regression models revealed significant interaction effects of intervention over time on non-adherence defined as percent of doses not-taken (IRR = 1.011, CI = 1.000-1.018) and percent of doses not-taken on time (IRR = 1.006, CI = 1.001-1.011) in the 30 days preceding each assessment. There were no significant differences between groups, but trends were observed for the MI-CBT/mDOT group to have greater 12 week on-time and worse 48 week adherence than the SC group. Findings of modest to null impact on adherence despite intensive interventions highlights the need for more effective interventions to maintain high adherence over time.

Development and efficacy of an intervention to enhance readiness for adherence among adults who had previously failed HIV treatment.

This paper outlines the development and initial testing of the READY intervention that was designed to enhance readiness for adherence among adults with a history of nonadherence to HIV treatment. Participants in this study were adults (n = 28) who ranged in age from 24 to 57: most were male (75%) and African American (64%). Participants had failed an average of four prior HIV treatment regimens due to nonadherence and were beginning a new regimen of protease inhibitor (PI)-based antiretroviral medications. The study was conducted from 2003 to 2006, prior to the standard use of boosted PI regimens. Results indicated that 50% of participants became adherent and had suppressed viral loads to less than 50 copies per milliliter at the 3-month postintervention follow-up time point. Of those who became adherent, 79% remained adherent at the 12-month postintervention follow-up time point. Implementation of the intervention was found to be feasible in a real-world setting and participants reported that they liked the intervention. A 6-session length of the intervention was found to have the same impact on adherence outcomes as a 12-session length. No differences were found in outcomes with regard to the intervention's start time: before or at the same time the new antiretroviral regimen was initiated. These results suggest that the READY intervention may have merit and that the 6-session length may be more acceptable. However, a larger controlled study is indicated to examine intervention efficacy further.