DOCSLIB.ORG

Speech and Language Processing

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Speech and Language Processing Speech and Language Processing Chapter 8 of SLP Speech Synthesis Outline 1) Arpabet 2) TTS Architectures 3) TTS Components • Text Analysis • Text Normalization • Homonym Disambiguation • Grapheme-to-Phoneme (Letter-to-Sound) • Intonation • Waveform Generation • Unit Selection • Diphones 13-03-05 Speech and Language Processing Jurafsky and Martin 2 Dave Barry on TTS “And computers are getting smarter all the time; scientists tell us that soon they will be able to talk with us. (By "they", I mean computers; I doubt scientists will ever be able to talk to us.) 13-03-05 Speech and Language Processing Jurafsky and Martin 3 ARPAbet Vowels b_d ARPA b_d ARPA 1 bead iy 9 bode ow 2 bid ih 10 booed uw 3 bayed ey 11 bud ah 4 bed eh 12 bird er 5 bad ae 13 bide ay 6 bod(y) aa 14 bowed aw 7 bawd ao 15 Boyd oy 8 Budd(hist) uh 13-03-05 Speech and Language Processing Jurafsky and Martin 4 Brief Historical Interlude • Pictures and some text from Hartmut Traunmüller’s web site: • http://www.ling.su.se/staff/hartmut/kemplne.htm • Von Kempeln 1780 b. Bratislava 1734 d. Vienna 1804 • Leather resonator manipulated by the operator to copy vocal tract configuration during sonorants (vowels, glides, nasals) • Bellows provided air stream, counterweight provided inhalation • Vibrating reed produced periodic pressure wave 13-03-05 Speech and Language Processing Jurafsky and Martin 5 Von Kempelen: • Small whistles controlled consonants • Rubber mouth and nose; nose had to be covered with two fingers for non-nasals • Unvoiced sounds: mouth covered, auxiliary bellows driven by string provides puff of air From Traunmüller’s web site 13-03-05 Speech and Language Processing Jurafsky and Martin 6 Modern TTS systems . 1960’s first full TTS: Umeda et al (1968) . 1970’s . Joe Olive 1977 concatenation of linear-prediction diphones . Speak and Spell . 1980’s . 1979 MIT MITalk (Allen, Hunnicut, Klatt) . 1990’s-present . Diphone synthesis . Unit selection synthesis 13-03-05 Speech and Language Processing Jurafsky and Martin 7 2. Overview of TTS: Architectures of Modern Synthesis . Articulatory Synthesis: . Model movements of articulators and acoustics of vocal tract . Formant Synthesis: . Start with acoustics, create rules/filters to create each formant . Concatenative Synthesis: . Use databases of stored speech to assemble new utterances. 13-03-05 Text from Richard Sproat slides Speech and Language Processing Jurafsky and Martin 8 Formant Synthesis . Were the most common commercial systems while computers were relatively underpowered. 1979 MIT MITalk (Allen, Hunnicut, Klatt) . 1983 DECtalk system . The voice of Stephen Hawking 13-03-05 Speech and Language Processing Jurafsky and Martin 9 Concatenative Synthesis . All current commercial systems. Diphone Synthesis . Units are diphones; middle of one phone to middle of next. Why? Middle of phone is steady state. Record 1 speaker saying each diphone . Unit Selection Synthesis . Larger units . Record 10 hours or more, so have multiple copies of each unit . Use search to find best sequence of units 13-03-05 Speech and Language Processing Jurafsky and Martin 10 TTS Demos (all are Unit-Selection) . Festival . http://www-2.cs.cmu.edu/~awb/festival_demos/index.html . Cepstral . http://www.cepstral.com/cgi-bin/demos/general . IBM . http://www-306.ibm.com/software/pervasive/tech/demos/tts.shtml 13-03-05 Speech and Language Processing Jurafsky and Martin 11 Architecture . The three types of TTS . Concatenative . Formant . Articulatory . Only cover the segments+f0+duration to waveform part. A full system needs to go all the way from random text to sound. 13-03-05 Speech and Language Processing Jurafsky and Martin 12 Two steps . PG&E will file schedules on April 20. TEXT ANALYSIS: Text into intermediate representation: . WAVEFORM SYNTHESIS: From the intermediate representation into waveform 13-03-05 Speech and Language Processing Jurafsky and Martin 13 The Hourglass 13-03-05 Speech and Language Processing Jurafsky and Martin 14 1. Text Normalization . Analysis of raw text into pronounceable words: . Sentence Tokenization . Text Normalization . Identify tokens in text . Chunk tokens into reasonably sized sections . Map tokens to words . Identify types for words 13-03-05 Speech and Language Processing Jurafsky and Martin 15 Rules for end-of-utterance detection . A dot with one or two letters is an abbrev . A dot with 3 cap letters is an abbrev. An abbrev followed by 2 spaces and a capital letter is an end-of-utterance . Non-abbrevs followed by capitalized word are breaks . This fails for . Cog. Sci. Newsletter . Lots of cases at end of line. Badly spaced/capitalized sentences 13-03-05 From Alan Black lecture notes Speech and Language Processing Jurafsky and Martin 16 Decision Tree: is a word end- of-utterance? 13-03-05 Speech and Language Processing Jurafsky and Martin 17 Learning Decision Trees . DTs are rarely built by hand . Hand-building only possible for very simple features, domains . Lots of algorithms for DT induction 13-03-05 Speech and Language Processing Jurafsky and Martin 18 Next Step: Identify Types of Tokens, and Convert Tokens to Words . Pronunciation of numbers often depends on type: . 1776 date: . seventeen seventy six. 1776 phone number: . one seven seven six . 1776 quantifier: . one thousand seven hundred (and) seventy six . 25 day: 13-03-05 . twenty-fifth Speech and Language Processing Jurafsky and Martin 19 Classify token into 1 of 20 types . EXPN: abbrev, contractions (adv, N.Y., mph, gov’t) . LSEQ: letter sequence (CIA, D.C., CDs) . ASWD: read as word, e.g. CAT, proper names . MSPL: misspelling . NUM: number (cardinal) (12,45,1/2, 0.6) . NORD: number (ordinal) e.g. May 7, 3rd, Bill Gates II . NTEL: telephone (or part) e.g. 212-555-4523 . NDIG: number as digits e.g. Room 101 . NIDE: identifier, e.g. 747, 386, I5, PC110 . NADDR: number as stresst address, e.g. 5000 Pennsylvania . NZIP, NTIME, NDATE, NYER, MONEY, BMONY, PRCT,URL,etc . SLNT: not spoken (KENT*REALTY) 13-03-05 Speech and Language Processing Jurafsky and Martin 20 More about the types . 4 categories for alphabetic sequences: . EXPN: expand to full word or word seq (fplc for fireplace, NY for New York) . LSEQ: say as letter sequence (IBM) . ASWD: say as standard word (either OOV or acronyms) . 5 main ways to read numbers: . Cardinal (quantities) . Ordinal (dates) . String of digits (phone numbers) . Pair of digits (years) . Trailing unit: serial until last non-zero digit: 8765000 is “eight seven six five thousand” (some phone numbers, long addresses) . But still exceptions: (947-3030, 830-7056) 13-03-05 Speech and Language Processing Jurafsky and Martin 21 Finally: expanding NSW Tokens . Type-specific heuristics . ASWD expands to itself . LSEQ expands to list of words, one for each letter . NUM expands to string of words representing cardinal . NYER expand to 2 pairs of NUM digits… . NTEL: string of digits with silence for puncutation . Abbreviation: . use abbrev lexicon if it’s one we’ve seen . Else use training set to know how to expand . Cute idea: if “eat in kit” occurs in text, “eat-in 13-03-05 kitchen” will also occur somewhere.Speech and Language Processing Jurafsky and Martin 22 2. Homograph disambiguation 19 most frequent homographs, from Liberman and Church use 319 survey 91 increase 230 project 90 close 215 separate 87 record 195 present 80 house 150 read 72 contract 143 subject 68 lead 131 rebel 48 live 130 finance 46 lives 105 estimate 46 protest 94 Not a huge problem, but still important 13-03-05 Speech and Language Processing Jurafsky and Martin 23 POS Tagging for homograph disambiguation . Many homographs can be distinguished by POS . use y uw s y uw z . close k l ow s k l ow z . house h aw s h aw z . live l ay v l ih v . REcord reCORD . INsult inSULT . OBject obJECT . OVERflow overFLOW . DIScount disCOUNT . CONtent conTENT 13-03-05 Speech and Language Processing Jurafsky and Martin 24 3. Letter-to-Sound: Getting from words to phones . Two methods: . Dictionary-based . Rule-based (Letter-to-sound=LTS) . Early systems, all LTS . MITalk was radical in having huge 10K word dictionary . Now systems use a combination 13-03-05 Speech and Language Processing Jurafsky and Martin 25 Pronunciation Dictionaries: CMU . CMU dictionary: 127K words . http://www.speech.cs.cmu.edu/cgi-bin/cmudict . Some problems: . Has errors . Only American pronunciations . No syllable boundaries . Doesn’t tell us which pronunciation to use for which homophones . (no POS tags) . Doesn’t distinguish case . The word US has 2 pronunciations . [AH1 S] and [Y UW1 EH1 S] 13-03-05 Speech and Language Processing Jurafsky and Martin 26 Pronunciation Dictionaries: UNISYN . UNISYN dictionary: 110K words (Fitt 2002) . http://www.cstr.ed.ac.uk/projects/unisyn/ . Benefits: . Has syllabification, stress, some morphological boundaries . Pronunciations can be read off in . General American . RP British . Australia . Etc . (Other dictionaries like CELEX not used because too small, British-only) 13-03-05 Speech and Language Processing Jurafsky and Martin 27 Dictionaries aren’t sufficient . Unknown words (= OOV = “out of vocabulary”) . Increase with the (sqrt of) number of words in unseen text . Black et al (1998) OALD on 1st section of Penn Treebank: . Out of 39923 word tokens, . 1775 tokens were OOV: 4.6% (943 unique types): names unknown Typos/other 1360 351 64 76.6% 19.8% 3.6% . So commercial systems have 4-part system: . Big dictionary . Names handled by special routines . Acronyms handled by special routines (previous lecture) . 13-03-05 Machine learned g2p algorithm for otherSpeech andunknown Language Processing Jurafsky and Martin 28 words Names . Big problem area is names . Names are common . 20% of tokens in typical newswire text will be names . 1987 Donnelly list (72 million households) contains about 1.5 million names .
Load more

Recommended publications
  • Rule-Based Standard Arabic Phonetization at Phoneme, Allophone, and Syllable Level
    Fadi Sindran, Firas Mualla, Tino Haderlein, Khaled Daqrouq & Elmar Nöth Rule-Based Standard Arabic Phonetization at Phoneme, Allophone, and Syllable Level Fadi Sindran [email protected] Faculty of Engineering Department of Computer Science /Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Firas Mualla [email protected] Faculty of Engineering Department of Computer Science /Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Tino Haderlein [email protected] Faculty of Engineering Department of Computer Science/Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Khaled Daqrouq [email protected] Department of Electrical and Computer Engineering King Abdulaziz University, Jeddah, 22254, Saudi Arabia Elmar Nöth [email protected] Faculty of Engineering Department of Computer Science /Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Abstract Phonetization is the transcription from written text into sounds. It is used in many natural language processing tasks, such as speech processing, speech synthesis, and computer-aided pronunciation assessment. A common phonetization approach is the use of letter-to-sound rules developed by linguists for the transcription from grapheme to sound. In this paper, we address the problem of rule-based phonetization of standard Arabic. 1The paper contributions can be summarized as follows: 1) Discussion of the transcription rules of standard Arabic which were used in literature on the phonemic and phonetic level. 2) Improvements of existing rules are suggested and new rules are introduced. Moreover, a comprehensive algorithm covering the phenomenon of pharyngealization in standard Arabic is proposed.
    [Show full text]
  • Lecture # 07 (Phonetics)
    Phonetics Phonetics (Introduction), Syllables, - text-to-speech recognition, - Onset, nucleus - Phones, - Coda, Rime, Speech sounds and Phonetic - Consonants and vowels, transcription, Phonological Categories and - International Phonetic Alphabet Pronunciation variation, (IPA), - ARPAbet, - Phonetic features, Articulatory Phonetics, - Predicting Phonetic variation, - The Vocal organs, - Factors Influencing Phonetic - Consonants: Place of Articulation, variation, - Consonants: Manner of Articulation, - Vowels, - Syllables, @Copyrights: Natural Language Processing (NLP) Organized by Dr. Ahmad Jalal (http://portals.au.edu.pk/imc/) 1. Phonetics (Introduction) “ What is the proper pronunciation of word “Special”” Text-to-speech conversation (converting strings of text words into acoustic waveforms). “Phonics” methods of teaching to children; - is first glance like a purely modern educational debate. Phonetics is the study of linguistic sounds, - how they are produced by the articulators of the human vocal tract, - how they are realized acoustically, - and how this acoustic realization can be digitized and processed. A key element of both speech recognition and text-to-speech systems: - how words are pronounced in terms of individual speech units called phones. @Copyrights: Natural Language Processing (NLP) Organized by Dr. Ahmad Jalal (http://portals.au.edu.pk/imc/) 2. Speech Sounds and Phonetic Transcription The study of the pronunciation of words is part of the field of phonetics, We model the pronunciation of a word as a string of symbols which represent phones or segments. - A phone is a speech sound; - phones are represented with phonetic symbols that bear some resemblance to a letter. We use 2 different alphabets for describing phones as; (1) The International Phonetic Alphabet (IPA) - is an evolving standard originally developed by the International Phonetic Association in 1888 with the goal of transcribing the sounds of all human languages.
    [Show full text]
  • UC Berkeley Dissertations, Department of Linguistics
    UC Berkeley Dissertations, Department of Linguistics Title Relationship between Perceptual Accuracy and Information Measures: A cross-linguistic Study Permalink https://escholarship.org/uc/item/7tx5t8bt Author Kang, Shinae Publication Date 2015 eScholarship.org Powered by the California Digital Library University of California Relationship between perceptual accuracy and information measures: A cross-linguistic study by Shinae Kang A dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Linguistics in the Graduate Division of the University of California, Berkeley Committee in charge: Professor Keith A. Johnson, Chair Professor Sharon Inkelas Professor Susan S. Lin Professor Robert T. Knight Fall 2015 Relationship between perceptual accuracy and information measures: A cross-linguistic study Copyright 2015 by Shinae Kang 1 Abstract Relationship between perceptual accuracy and information measures: A cross-linguistic study by Shinae Kang Doctor of Philosophy in Linguistics University of California, Berkeley Professor Keith A. Johnson, Chair The current dissertation studies how the information conveyed by different speech el- ements of English, Japanese and Korean correlates with perceptual accuracy. Two well- established information measures are used: weighted negative contextual predictability (in- formativity) of a speech element; and the contribution of a speech element to syllable differ- entiation, or functional load. This dissertation finds that the correlation between information and perceptual accuracy differs depending on both the type of information measure and the language of the listener. To compute the information measures, Chapter 2 introduces a new corpus consisting of all the possible syllables for each of the three languages. The chapter shows that the two information measures are inversely correlated.
    [Show full text]
  • Multilingual Speech Recognition for Selected West-European Languages
    master’s thesis Multilingual Speech Recognition for Selected West-European Languages Bc. Gloria María Montoya Gómez May 25, 2018 advisor: Doc. Ing. Petr Pollák, CSc. Czech Technical University in Prague Faculty of Electrical Engineering, Department of Circuit Theory Acknowledgement I would like to thank my advisor, Doc. Ing. Petr Pollák, CSc. I was very privileged to have him as my mentor. Pollák was always friendly and patient to give me his knowl- edgeable advise and encouragement during my study at Czech Technical University in Prague. His high standards on quality taught me how to do good scientific work, write, and present ideas. Declaration I declare that I worked out the presented thesis independently and I quoted all used sources of information in accord with Methodical instructions about ethical principles for writing academic thesis. iii Abstract Hlavním cílem předložené práce bylo vytvoření první verze multilingválního rozpozná- vače řeči pro vybrané 4 západoevropské jazyky. Klíčovým úkolem této práce bylo de- finovat vztahy mezi subslovními akustickými elementy napříč jednotlivými jazyky při tvorbě automatického rozpoznávače řeči pro více jazyků. Vytvořený multilingvální sys- tém pokrývá aktuálně následující jazyky: angličtinu, němčinu, portugalštinu a španěl- štinu. Jelikož dostupná fonetická reprezentace hlásek pro jednotlivé jazyky byla různá podle použitých zdrojových dat, prvním krokem této práce bylo její sjednocení a vy- tvoření sdílené fonetické reprezentace na bázi abecedy X-SAMPA. Pokud jsou dále acoustické subslovní elementy reprezentovány sdílenými skrytými Markovovy modely, případný nedostatek zdrojových dat pro trénováni může být pokryt z jiných jazyků. Dalším krokem byla vlastní realizace multilingválního systému pomocí nástrojové sady KALDI. Použité jazykové modely byly na bázi zredukovaných trigramových modelů zís- kaných z veřejně dostupých zdrojů.
    [Show full text]
  • Phonemic Similarity Metrics to Compare Pronunciation Methods
    Phonemic Similarity Metrics to Compare Pronunciation Methods Ben Hixon1, Eric Schneider1, Susan L. Epstein1,2 1 Department of Computer Science, Hunter College of The City University of New York 2 Department of Computer Science, The Graduate Center of The City University of New York [email protected], [email protected], [email protected] error rate. The minimum number of insertions, deletions and Abstract substitutions required for transformation of one sequence into As grapheme­to­phoneme methods proliferate, their careful another is the Levenshtein distance [3]. Phoneme error rate evaluation becomes increasingly important. This paper ex­ (PER) is the Levenshtein distance between a predicted pro­ plores a variety of metrics to compare the automatic pronunci­ nunciation and the reference pronunciation, divided by the ation methods of three freely­available grapheme­to­phoneme number of phonemes in the reference pronunciation. Word er­ packages on a large dictionary. Two metrics, presented here ror rate (WER) is the proportion of predicted pronunciations for the first time, rely upon a novel weighted phonemic substi­ with at least one phoneme error to the total number of pronun­ tution matrix constructed from substitution frequencies in a ciations. Neither WER nor PER, however, is a sufficiently collection of trusted alternate pronunciations. These new met­ sensitive measurement of the distance between pronunciations. rics are sensitive to the degree of mutability among phonemes. Consider, for example, two pronunciation pairs that use the An alignment tool uses this matrix to compare phoneme sub­ ARPAbet phoneme set [4]: stitutions between pairs of pronunciations. S OW D AH S OW D AH Index Terms: grapheme­to­phoneme, edit distance, substitu­ S OW D AA T AY B L tion matrix, phonetic distance measures On the left are two reasonable pronunciations for the English word “soda,” while the pair on the right compares a pronuncia­ 1.
    [Show full text]
  • Phonetic Properties of Oral Stops in Three Languages with No Voicing Distinction
    City University of New York (CUNY) CUNY Academic Works All Dissertations, Theses, and Capstone Projects Dissertations, Theses, and Capstone Projects 9-2018 Phonetic Properties of Oral Stops in Three Languages with No Voicing Distinction Stephanie M. Kakadelis The Graduate Center, City University of New York How does access to this work benefit ou?y Let us know! More information about this work at: https://academicworks.cuny.edu/gc_etds/2871 Discover additional works at: https://academicworks.cuny.edu This work is made publicly available by the City University of New York (CUNY). Contact: [email protected] PHONETIC PROPERTIES OF ORAL STOPS IN THREE LANGUAGES WITH NO VOICING DISTINCTION by STEPHANIE MARIE KAKADELIS A dissertation submitted to the Graduate Faculty in Linguistics in partial fulfillment of the requirements for the degree of Doctor of Philosophy, The City University of New York 2018 © 2018 STEPHANIE MARIE KAKADELIS All Rights Reserved ii Phonetic Properties of Oral Stops in Three Languages with No Voicing Distinction by Stephanie Marie Kakadelis This manuscript has been read and accepted for the Graduate Faculty in Linguistics in satisfaction of the dissertation requirement for the degree of Doctor of Philosophy. Date Juliette Blevins Chair of Examining Committee Date Gita Martohardjono Executive Officer Supervisory Committee: Douglas H. Whalen Jason Bishop Claire Bowern (Yale University) THE CITY UNIVERSITY OF NEW YORK iii ABSTRACT Phonetic Properties of Oral Stops in Three Languages with No Voicing Distinction by Stephanie Marie Kakadelis Advisor: Juliette Blevins Almost all studies on the phonetics of oral stop voicing patterns focus on languages with a voicing distinction. This gives rise to some debate regarding which aspects of voicing patterns arise from inherent articulatory effects related to the production of a voicing distinction, and which aspects are intentional adjustments by speakers meant to enhance a phonological contrast.
    [Show full text]
  • Techniques and Challenges in Speech Synthesis Final Report for ELEC4840B
    Techniques and Challenges in Speech Synthesis Final Report for ELEC4840B David Ferris - 3109837 04/11/2016 A thesis submitted in partial fulfilment of the requirements for the degree of Bachelor of Engineering in Electrical Engineering at The University of Newcastle, Australia. Abstract The aim of this project was to develop and implement an English language Text-to-Speech synthesis system. This first involved an extensive study of the mechanisms of human speech production, a review of modern techniques in speech synthesis, and analysis of tests used to evaluate the effectiveness of synthesized speech. It was determined that a diphone synthesis system was the most effective choice for the scope of this project. A diphone synthesis system operates by concatenating sections of recorded human speech, with each section containing exactly one phonetic transition. By using a database that contains recordings of all possible phonetic transitions within a language, or diphones, a diphone synthesis system can produce any word by concatenating the correct diphone sequence. A method of automatically identifying and extracting diphones from prompted speech was designed, allowing for the creation of a diphone database by a speaker in less than 40 minutes. The Carnegie Mellon University Pronouncing Dictionary, or CMUdict, was used to determine the pronunciation of known words. A system for smoothing the transitions between diphone recordings was designed and implemented. CMUdict was then used to train a maximum-likelihood prediction system to determine the correct pronunciation of unknown English language alphabetic words. Using this, the system was able to find an identical or reasonably similar pronunciation for over 76% of the training set.
    [Show full text]
  • Automatic Phonetization-Based Statistical Linguistic Study of Standard Arabic
    Fadi Sindran, Firas Mualla, Tino Haderlein, Khaled Daqrouq & Elmar Nöth Automatic Phonetization-based Statistical Linguistic Study of Standard Arabic Fadi Sindran [email protected] Faculty of Engineering/ Department of Computer Science Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Firas Mualla [email protected] Faculty of Engineering/ Department of Computer Science Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Tino Haderlein [email protected] Faculty of Engineering/ Department of Computer Science Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Khaled Daqrouq [email protected] Department of Electrical and Computer Engineering King Abdulaziz University, Jeddah, 22254, Saudi Arabia Elmar Nöth [email protected] Faculty of Engineering/ Department of Computer Science Pattern Recognition Lab Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen, 91058, Germany Abstract Statistical studies based on automatic phonetic transcription of Standard Arabic texts are rare, and even though studies have been performed, they have been done only on one level – phoneme or syllable – and the results cannot be generalized on the language as a whole. In this paper we automatically derived accurate statistical information about phonemes, allophones, syllables, and allosyllables in Standard Arabic. A corpus of more than 5 million words, including words and sentences
    [Show full text]
  • A Tutorial on Acoustic Phonetic Feature Extraction for Automatic Speech Recognition (ASR) and Text-To-Speech (TTS) Applications in African Languages
    Linguistic Portfolios Volume 9 Article 11 2020 A Tutorial on Acoustic Phonetic Feature Extraction for Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) Applications in African Languages Ettien Koffi St. Cloud State University, [email protected] Follow this and additional works at: https://repository.stcloudstate.edu/stcloud_ling Part of the Applied Linguistics Commons Recommended Citation Koffi, Ettien (2020)A " Tutorial on Acoustic Phonetic Feature Extraction for Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) Applications in African Languages," Linguistic Portfolios: Vol. 9 , Article 11. Available at: https://repository.stcloudstate.edu/stcloud_ling/vol9/iss1/11 This Article is brought to you for free and open access by theRepository at St. Cloud State. It has been accepted for inclusion in Linguistic Portfolios by an authorized editor of theRepository at St. Cloud State. For more information, please contact [email protected]. Koffi: A Tutorial on Acoustic Phonetic Feature Extraction for Automatic Linguistic Portfolios–ISSN 2472-5102 –Volume 9, 2020 | 130 A TUTORIAL ON ACOUSTIC PHONETIC FEATURE EXTRACTION FOR AUTOMATIC SPEECH RECOGNITION (ASR) AND TEXT-TO-SPECH (TTS) APPLICATIONS IN AFRICAN LANGUAGES ETTIEN KOFFI ABSTRACT At present, Siri, Dragon Dictate, Google Voice, and Alexa-like functionalities are not available in any indigenous African language. Yet, a 2015 Pew Research found that between 2002 to 2014, mobile phone usage increased tenfold in Africa, from 8% to 83%.1 The Acoustic Phonetic Approach (APA) discussed in this paper lays the foundation that will make Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) applications possible in African languages. The paper is written as a tutorial so that others can use the information therein to help digitalize many of the continent’s indigenous languages.
    [Show full text]
  • Synchronizing Keyframe Facial Animation to Multiple Text-To- Speech Engines and Natural Voice with Fast Response Time
    Dartmouth College Dartmouth Digital Commons Dartmouth College Undergraduate Theses Theses and Dissertations 5-1-2004 Synchronizing Keyframe Facial Animation to Multiple Text-to- Speech Engines and Natural Voice with Fast Response Time William H. Pechter Dartmouth College Follow this and additional works at: https://digitalcommons.dartmouth.edu/senior_theses Part of the Computer Sciences Commons Recommended Citation Pechter, William H., "Synchronizing Keyframe Facial Animation to Multiple Text-to-Speech Engines and Natural Voice with Fast Response Time" (2004). Dartmouth College Undergraduate Theses. 38. https://digitalcommons.dartmouth.edu/senior_theses/38 This Thesis (Undergraduate) is brought to you for free and open access by the Theses and Dissertations at Dartmouth Digital Commons. It has been accepted for inclusion in Dartmouth College Undergraduate Theses by an authorized administrator of Dartmouth Digital Commons. For more information, please contact [email protected]. Synchronizing Keyframe Facial Animation to Multiple Text-to-Speech Engines and Natural Voice with Fast Response Time A Senior Honors Thesis Submitted to the faculty in partial fulfillment of the requirements for the Degree of Bachelor of Arts in Computer Science Dartmouth Computer Science Technical Report TR2004-501 by William H. Pechter Dartmouth College Hanover, NH May 2004 _________________ Lorie Loeb _________________ Hany Farid _________________ Stephen Linder i Abstract This thesis aims to create an automated lip-synchronization system for real-time applications. Specifically, the system is required to be fast, consist of a limited number of keyframes with small memory requirements, and create fluid and believable animations that synchronize with text-to-speech engines as well as raw voice data. The algorithms utilize traditional keyframe animation and a novel method of keyframe selection.
    [Show full text]
  • UC Santa Cruz UC Santa Cruz Electronic Theses and Dissertations
    UC Santa Cruz UC Santa Cruz Electronic Theses and Dissertations Title Development and Evaluation of Software Tools for Speech Therapy Permalink https://escholarship.org/uc/item/0rk63095 Author Rubin, Zachary Publication Date 2017 License https://creativecommons.org/licenses/by/4.0/ 4.0 Peer reviewed|Thesis/dissertation eScholarship.org Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA SANTA CRUZ DEVELOPMENT AND EVALUATION OF SOFTWARE TOOLS FOR SPEECH THERAPY A dissertation submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in COMPUTER ENGINEERING by Zachary Rubin March 2017 The Dissertation of Zachary Rubin is approved: Sri Kurniawan, Chair Travis Tollefson Mircea Teodorescu Tyrus Miller Vice Provost and Dean of Graduate Studies Copyright © by Zachary Rubin 2017 Table of Contents List of Figures v List of Tables vii Abstract viii Dedication x Acknowledgments xi 1 Introduction 1 1.1 Cleft Palate . .2 1.2 Speech Recognition . .6 2 Proof of Concept: Mobile Mispronunciation Recognition System 9 2.1 Related Work . 10 2.2 Requirements Gathering . 11 2.3 Development . 13 2.4 Lateral Lisp Test . 14 3 Pilot Test: Speech Adventure 19 3.1 Related Work . 20 3.2 Requirements Gathering . 21 3.2.1 User Interviews . 22 3.2.2 Video Game Analysis . 25 3.3 Development . 30 3.3.1 Video Game . 31 3.3.2 Speech Recognition . 36 3.3.3 Statistics . 41 3.4 In-Office Evaluation . 42 4 Long Term Test: Speech With Sam 46 4.1 Related Work . 47 4.2 Requirements Gathering . 49 4.3 Development .
    [Show full text]
  • Learning Allophones: What Input Is Necessary?
    LearningAllophones:WhatInputIsNecessary? Caitlin Richter 1. Introduction The discovery of phonemes and allophonic relations, as surface alternations accumulate in children’s growing vocabulary, is a vital element of language acqui- sition. This paper quantitatively models allophone learning, when the child’s input triggers a merger of surface-distinct segments into a single underlying phoneme, illustrated here for the case of grouping US English alveolar flaps [R] together with voiceless alveolar stops [t] as allophones of an underlying /T/. The Tolerance Principle (Yang, 2016), grounded in considerations of process- ing efficiency, defines for the general case the tipping point at which any grammar becomes unsustainable for a child, given the linguistic items that they know a grammatical rule should apply to and how many of them evidently violate it. Inte- grating this cognitively motivated model of grammar evaluation with a principled model for the nature of the phonological system the child will be evaluating at some non-final acquisitional stage (Ringe & Eska, 2013), it is possible to trace a portion of the development of the phonological inventory, as it is continually re- evaluated with growing vocabulary and potentially ultimately rejected in favour of another grammar. Therefore, the present model of allophone acquisition proposes firstly that the learner’s default initial grammar is roughly ‘what you hear is what you get’ in terms of the language’s surface contrasts, as specified by Invariant Transparency (an acquisitional hypothesis soundly implied by patterns of histor- ical language change, §1.1; Ringe & Eska, 2013); and secondly, that this initial grammar is subject to the Tolerance Principle, such that if the child has too much violating input in which it is obvious that what is heard on the surface is not what is underlying then the initial grammar will become unsustainable.
    [Show full text]