130 likes | 144 Views
Outline. Introduction Methods Problems to be solved Demos. Speech Assessment. Speech assessment: How to assess an utterance for the purpose of learning a spoken language? Assessment levels: syllables, words, sentences, paragraphs
E N D
Outline • Introduction • Methods • Problems to be solved • Demos
Speech Assessment • Speech assessment: How to assess an utterance for the purpose of learning a spoken language? • Assessment levels: syllables, words, sentences, paragraphs • Assessment criteria: timbre, tone, energy, rhythm, co-articulation, … • Feedbacks: High-level correction and suggestions
Related Disciplines • Related disciplines for speech assessment: • Language learning: • CALL: Computer Assisted Language Learning • CAPT: Computer Assisted Pronunciation Training • Speech technology: • UV: Utterance Verification
Our Approach • Basic approach to timbre assessment • Lexicon net construction (Usually a sausage net) • Forced alignment to identify phone boundaries • Phone scoring based on several criteria, such as ranking, histograms, posterior prob., etc. • Weighted average to get syllable score • Weighted average to get sentence score
Basic Assessment Criteria • Timber • Based on acoustic models • Tone • Based on tone recognition (for tonal language) • Based on pitch similarity with the target utterance • Energy • Based on energy comparison with the target utterance • Rhythm • Based on duration comparison with the target utterance • Fluency
Additional Assessment Criteria • English • Stress • Levels (word or sentence) • Meanings • Intonation • Declarative sentence • Interrogative sentence • Co-articulation • A red apple. • Did you call me? • Hit and run • Mandarin • Tone • Retroflex or not • Co-articulation • 兒化音
Problems to be Solved • Score related • Optimization • Consistency • Interpretability • Confusing phone id. (日本人的發音) • Slightly adaptation • Paragraph-level assessment • Contents construction
Demo: Practice of Mandarin Idioms of Length 4 (一語中的) • Level (difficulty) of an idiom is based on it’s freq. via Google search: • 孤掌難鳴 ===> 260,000 • 鶼鰈情深 ===> 43,300 • 亡鈇意鄰 ===> 22,700 • 舉案齊眉 ===> 235,000 • Can be adapted for English learning • Next step: multi-threading, fast decoding via FSM
Support Mandarin & English Support user-defined recitation script Next step: multithreading for recording & recognition Demo: Recitation Machine(唸唸不忘)
Dialog-based practice and evaluation Demo: Dialog Practice via Videos
Demos on PC and PMP • PC軟體 • Lucy’s Café: Speech and Score • PMP • 華語練習機
Demo: Embedded Systems • Chicken run (落跑雞) • Penguin for Tang Poetry (唐詩企鵝) • Robot Fighter (蘿蔔戰士) • Singing Bass & Dog (大嘴鱸魚和唱歌狗)
On-going Work • On-going work: • Tone recognition and assessment • Retroflex & nonretroflex recognition • Detection of “兒化音” • Demo page: • http://mirlab.org/mir_main/demo.htm