🚨 New Paper Alert 🚨 🗣️ 📄 "How does a credible voice sound?" in the Journal of Acoustical Society of America @acousticsorg, featured on the cover of the May issue! 🔗 Read the paper here ( https://t.co/5Hl11yDmp6) #VoicePerception #Credibility #Acoustics #SpeechScience🧵1/5

Safe to say I am late to the academic Twitter world... In that spirit, here is a 'oldie' from 2024 where we first document the length of different types of stutters in a stuttered speech corpus from with with @tang_kevin & @UCL_SpeechResGr https://t.co/G1pV0TSbow

These insights enhance our understanding of how acoustic cues influence the perception of credibility in speech. They have potential applications in communication strategies🗣️, media presentations 🎥, and the development of AI voice systems🤖💬. #CommunicationResearch 🧵5/5

Key findings: Credible speech shows 🎤 higher energy (MFCC1) and 🗣️ faster rates than neutral. Irony features 🗣️ faster rates and higher spectral centroid. 👩‍🦰Gender differences in irony: women speak faster, men use higher pitch. 🧵4/5

We analyzed acoustic features distinguishing credible, neutral & ironic speech using a German corpus by amateur speakers. A machine-learning-based multinomial logistic regression with recursive feature elimination identified key features linked to perceived credibility. 🧵3/5

This is the first publication from the collaboration between my lab @SLaMLab_HHU ( https://t.co/Cqj00bPTeX) at @HHU_de and the Institute of Sound and Vibration Engineering, Hochschule Düsseldorf with Prof. Dr. Jochen Steffens, Patrick Blättermann and Maximilian Sattler. 🧵2/5

🚨New article alert for phonologists and phoneticians! Check out "A unified model of lenition as modulation reduction: gauging consonant strength in Ibibio" by John Harris, Eno-AbasiUrua & @tang_kevin , published in Phonology (2024). https://t.co/DqigrJ7mEX 🧵[1/4] #Lenition

📍Schadowplatz und im Haus der Universität 🛜 https://t.co/FI6IAENGwp @HHU_de Organization Team: Dalia Rodrigues, Vittorio Ciccarelli, @tang_kevin und @SLaMLab_HHU

Habt ihr euch schon einmal gefragt, wie Sprache Einblicke in Gesundheit, Gesellschaft oder politische Einstellungen geben kann? Und wie Studierende durch ihre eigene Forschung wertvolle Erfahrungen sammeln und neue Fähigkeiten entwickeln? Virtualstand:

Das Speech, Lexicon and Modeling Lab (@SLaMLab_HHU,@tang_kevin) lädt euch ein, spannende Einblicke in die Welt der Sprachwissenschaft zu gewinnen! Besucht uns auf der #NachtderWissenschaft am 13.09.24 (Standnummber 9) und entdeckt die faszinierenden Projekte unserer Studierenden

Could computational methods like Phonet be used as an additional or alternative method of lenition measurement? https://t.co/AYshUldlsg #acoustics #speech @tang_kevin,@SLaMLab_HHU @UFLinguistics,@UFCISE @HHU_de

At @ISCAInterspeech? Do check out our work tomorrow at 11:20 on modeling probabilistic reduction across linguistic domains with Naive Discriminative Learning. Work by @_ansost_ based on her BA thesis and @tang_kevin. Paper: https://t.co/4k5A8VQ5E6 @UFLinguistics, @HHU_de @IL_hhu

Authors: Kevin Tang @tang_kevin @HHU_de, Ratree Wayland @UFLinguistics, Fenqi Wang @FenqiW @SFULinguistics and Sophia Vellozzi @LinkedIn and Rahul Sengupta @UFCISE Paper: https://t.co/JqsC6uDGrp Funded by @NSF (2037266). #NeuralNetworks #Spanish #Weakening #Phonetics 🧵6/6

Lastly 🛠️ we’ve made Phonet more accessible by releasing the trained Spanish Phonet model and a step-by-step pipeline for training and inferring new models. Ready to explore? Link to the tool: https://t.co/sujV2aAiww 🌐🔧 #OpenScience #LinguisticsTools 🧵5/6

🎯 This study highlights the effectiveness of using neural network that is phonologically informed as an alternative or supplementary method for measuring lenition. The consistency is promising for future research in speech science and phonological theories! ✨ 🧵4/6

🔄 While traditional acoustic metrics gave mixed and inconsistent results, Phonet's measures shined! 🥇🌟 Sonorant and continuant probabilities aligned with expected lenition patterns, influenced by voicing, articulation, and context. 🔍 #InterpretableAI 🧵3/6

🔍 We compared three acoustic measures: 1⃣ Min & Max Intensity Velocity 📈 (Kingston 2008) 2⃣ Duration ⏳ 3⃣ Neural network (Phonet) measures based on the posterior probabilities of Sonorant & Continuant features 🤖 #MachineLearning 🧵2/6

🚀 Our new publication in @ASA_JASA evaluated the consistency of lenition (consonant weakening) measures, focusing on voiceless and voiced stops in Spanish & pre-released a linguistic toolkit for lenition research 🎙️🧠 #Phonetics #Lenition. https://t.co/i00mJaWsjh 🧵1/6

This study investigated the effects of Parkinson's disease and various linguistic factors on the degree of lenition in Spanish stops: https://t.co/N9IR3HRGeR #acoustics #speech @UFLinguistics @tang_kevin @SLaMLab_HHU @SFULinguistics @UFCISE

Linguists at @HHU_de (@IL_hhu @SLaMLab_HHU) are well-represented at #LabPhon19 in Seoul, Korea: @BraveMoneyLute, @tang_kevin, @_ansost_, @Akhilesh_k_r, @dmncschmtz and @dinahwonders. Together they will present three papers. See links for abstracts! 🧵unroll [1/4]