DOCSLIB.ORG

Tool for Czech Pronunciation Generation Combining Fixed Rules with Pronunciation Lexicon and Lexicon Management Tool

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Tool for Czech Pronunciation Generation Combining Fixed Rules with Pronunciation Lexicon and Lexicon Management Tool Tool for Czech Pronunciation Generation Combining Fixed Rules with Pronunciation Lexicon and Lexicon Management Tool Petr PollaÂk, VaÂclav HanzÏl Czech Technical University in Prague CVUT FEL K331 Technicka 2, 166 27 Praha 6 - Dejvice, Czech Republic [email protected], [email protected] Abstract This paper presents two different tools which may be used as a support of speech recognition. The tool ªtranscº is the ®rst one and it generates the phonetic transcription (pronunciation) of given utterance. It is based mainly on ®xed rules which can be de®ned for Czech pronunciation but it can work also with speci®ed list of exceptions which is de®ned on lexicon basis. It allows the usage of ªtranscº for unknown text with high probability of correct phonetic transcription generation. The second part is devoted to lexicon management tool ªlexeditº which may be useful in the phase of generation of pronunciation lexicon for collected corpora. The presented tool allows editing of pronunciation, playing examples of pronunciation, comparison with reference lexicon, updating of reference lexicon, etc. 1. Introduction 2. Phonetical alphabet for Czech One of the most important tasks within speech reco- It is fact that any of®cially de®ned standard of Czech gnition is the generation of the pronunciation for given phoneme inventory has not been de®ned yet. We can ®nd utterance. Both in training or in recognition phase, the in- several attempts to its de®nition, e.g. in (Nouza et al., formation about the pronunciation of given or supposed 1997) or (http://www.phon.ucl.ac.uk/home/sampa/czech- utterance must be available. The pronunciation can be obta- uni.htm), but any alphabet has not been accepted ined either from a pronunciation lexicon or using a tool for as the standard. The latest Czech SAMPA proposal automated conversion of orthographic transcription to the is described in (http://noel.feld.cvut.cz/sampa/CZECH- phonetic one. This paper is devoted to the solution of this SAMPA.html) That's reason why we must brie¯y describe problem for Czech language. the alphabets used for phonetic transcription in this work. Czech has quite regular pronunciation, so the rules for We work with two different phonetical alphabets which automated generation of phonetic transcription (pronunci- differ only in used symbols for de®ned inventory of phone- ation) can be de®ned excepting some words, mainly the mes with only one exception. It will be discussed later. words of foreign origin. A special tool solving this pro- The ®rst one is IPA CTU (Internal Phonetic Alphabet blem ªtranscº was developed in Speech Processing Group created at Czech Technical University in Prague) which at Czech Technical University in Prague. When the automa- is very close to the alphabet published in (Nouza et al., ted generation of phonetic transcription cannot be applied 1997). It is used historically by many tools written at our for a word with some pronunciation irregularities (due a laboratory and it seems to be very convenient especially con¯ict with the general rule), the special simple syntax for the following reasons. Each phoneme is always repre- of orthographic transcription can be used to re-de®ne au- sented by single character and the symbols are very close tomatically generated regular pronunciation. This syntax is to the transcription of the pronunciation using standard described in section 3. Czech orthography. Consequently, it mainly respects stan- The second approach of utterance pronunciation gene- dard Czech pronunciation rules, i.e. applying ªtranscº to ration is based on pronunciation lexicon. This approach is the word in this phonetic transcription we obtain the same not optimal for Czech due to high number of in¯ections result. The exception from this rule can be ®nd for spe- in the language because it leads to growing size of the le- cial phonemes which are represented by upper case letters. xicon. On the other hand, irregular pronunciation can be These phonemes are quite rare and they may be represen- easily involved in the lexicon and any special orthography ted usually by equivalents with two phonemes written by transcription isn't required in this case. This is the main lower case letters (ªE - euº, ªA - auº, ªZ - dzº, etc.), see advantage of the lexicon approach. Because the creation of tab. 1. Many times it can be done because the threshold be- pronunciation lexicon is really hard work with necessary tween these two pronunciation is often quite soft, i.e. we can manual checks we have designed lexicon management tool observe smooth passing of two phonemes ªeuº into single ªlexeditº with quite simple control which is described in the phoneme ªEº. section 4. Finally, we use small lexicon as a part of tool for auto- As the ®rst step to standardization, CZECH SPEECH- mated pronunciation generation (ªtranscº), where the words DAT SAMPA was de®ned. It originates from IPA CTU but with pronunciation exceptions are stored. it uses already existing symbols in SAMPA alphabets for equivalent phonemes from other languages. This alphabet were de®ned within SpeechDat(E) project, see (CÏernocky et al., 2000). The correspondence between these two alphabets is in the tab. 1. There is only one small difference in pho- neme inventory. CZECH SPEECHDAT SAMPA doesn't of context sensitive grammars (which permits any order of resolve voiced and unvoiced allophones of ªrϺ. rules) but leads to much reduced complexity and also ena- CZECH SPEECHDAT SAMPA is the alphabet which bles us to use smaller set of symbols ± each 8-bit character is usually used at the international level. But the transcrip- may serve as both terminal and non-terminal symbol with tion of Czech pronunciations is very strange. That is reason slightly different meanings in various stages of processing. why for our private purposes we use IPA CTU. The Czech reader may appreciate very convenient phonetical transcrip- 3.1. Input format tion which is very close to Czech orthography and which is The orthographic transcription at the input of this generally easily understandable to Czech people, see tab 1. tool must obey conventions which can be summarized as When we annotated Czech SpeechDat database on the pro- follows: nunciation level we made the experience that people without any special education in this area could easily understand The convertible text must be written in regular words this transcription after short training period. with standard Czech syntax. The transcription can con- tain punctuation. 3. transc - tool for pronunciation generation The words with irregular pronunciation are typed with Czech language has generally regular pronunciation following syntax: ª(orthography/pronunciation)º. which corresponds closely with written form of the text. The pronunciation is typed using standard Czech Some irregularities appear mainly within the words of fore- alphabet in any way which yields the correct result ign origin. The most frequent case are personal and company according to regular pronunciation rules, i.e. according names, scienti®c terminology, technical terms, etc. Some of to rule ªwrite what you hearº. these irregularities are quite systematic and they can be handled by regular rules. Others can only be covered by Ex: (monitor/monytor) exception lexicon lookup. Czech, similarly as other Slavic languages, has highly in- Known irregularities are included in internal lexicon ¯ective nature. Typically, nouns, adjectives, pronouns, and of exceptions. This list can be extended by user de®ned numerals are declined. They may have up to 7 different external lexicon which includes transcriptions accor- forms both in plural and singular. Verbs may be conjugated ding to previously de®ned rule. The growing database in 6 different forms for each tense, sometimes the forms dif- of irregularities continuously decreases the probabi- fer according to gender. Using the pronunciation lexicon, its lity of bad orthographic-to-phonetic conversion, i.e. it size rapidly increases. Moreover, many forms of the words minimizes the necessity of special syntax in the ortho- differ only in terminations with regular pronunciation while graphy transcription of new utterances. their bases remain unchanged. That's the reason why the The transcription is generally case insensitive. Whole usage of ®xed pronunciation rules in a conversion algori- text is in the beginning converted to lower case form. thm seems to be more ef®cient for the generation of written However sometimes it may be useful to distinguish text pronunciation than the lexicon approach. between two words which differ just in the case The most complicated part of ªregularº pronunciation of the letters. These words may be included in generation is proper handling of voiced/voiceless assimilati- the external lexicon. Order of lexicon lookup and ons in consonant sequences. Consonant sequences may be conversion to lowercase is problematic to decide, quite long in Czech words and most often the sequence assi- each possibility has some disadvantages ± therefore milates to be either voiced or unvoiced as a whole. However we repeat the lookup both before and after the con- individual consonants behave in different ways during this version to lowercase to enable case sensitive handling assimilation ± they may be more or less dominant and force where needed and case insensitive handling otherwise. its neighbor to share voiced/voiceless quality with them, they may be more or less likely to be dominated, these re- Ex: Detektor rÏecÏi znacÏÂmeõ VAD. lations may differ in forward and backward directions and Detektor rÏecÏi znacÏõÂme ve Âa deÂ. may depend on context. Chains of assimilations may cross detektor rÏecÏy znacÏÂmey ve Âa de word boundaries. On the other hand, certain consonants may split consonant sequences into parts which differ in Ex: VyÂrobek ma mnoho vad. voiced/voiceless quality. Overall number of combinations VyÂrobek ma mnoho vad. is immense and software realization of these ªcommon and vyÂrobek ma mnoho vat simpleº assimilations gets rather complicated.
Load more

Recommended publications
  • Teaching Initial Reading and Writing at the Very Beginning of Compulsory School Attendance in the Czech Primary School Phdr
    FELA Initial Teaching in Europe 1st Symposium in Klagenfurt 2015 Teaching initial reading and writing at the very beginning of compulsory school attendance in the Czech Primary School PhDr. Marie Ernestová, Czech Republic 1 Introductory Facts • language typology: Czech is a flectional language • 7 cases in both singular and plural noun declination patterns and 6 forms in verbal conjugation (3 in the singular and 3 in the plural) • illogical grammar gender system (table=masculine, chair=feminine, window=neuter). 2 • language genealogy: Czech is a West Slavic language • with highly phonemic orthography it may be described as having regular spelling • the Czech alphabet consists of 42 letters (including the digraph ch, which is considered a single letter in Czech): • a á b c č d ď e é ě f g h ch i í j k l m n ň o ó p q r ř s š t ť u ú ů v w x y ý z ž 3 • the Czech orthographic system is diacritic. The háček /ˇ/ is added to standard Latin letters for expressing sounds which are foreign to the Latin language. The acute accent is used for long vowels. • There are two ways in Czech to write long [u:]: ú or ů. 4 • only one digraph: ´ch´ always pronounced the same, as /X/ as in the English loch. • All over Europe the ch-consonant combination makes problems as this ´/X/´ sound can be spelt in as many as eighteen ways: c, ć, ç,ċ,ĉ, č, ch, çh, ci, cs,cz,tch, tj,tš, tsch, tsi, tsj, tx and even k.
    [Show full text]
  • ABSTRACT Orthographic Reform and Language Planning in Russian
    ABSTRACT Orthographic Reform and Language Planning in Russian History Sylvia Conatser Segura Director: Adrienne Harris, Ph.D. An oft-neglected topic of book history and culture, both in histories of writing development and in cultural histories of Russia and Eastern Europe, is the rich history of orthographic reform and language planning in Russian and Slavic history. Even more neglected are the cultural, religious, and political drivers of language planning and the relationship of identity to writing systems. This thesis aims to correct this oversight by synthesizing histories on the development of written Russian until its codification by Peter the Great in the early 18th century as well as the language reforms accomplished under the Bolsheviks in the 20th century. It will present analysis on motivations for these two periods of language reform as well as impulses behind resistance thereof; the resulting evidence will demonstrate the close ties of political and cultural, particularly religious, identity to the use of written systems as illustrated by the relationship of Eastern Orthodoxy to the Cyrillic alphabet. APPROVED BY DIRECTOR OF HONORS THESIS _________________________________________ Dr. Adrienne Harris, Department of Modern Languages and Culture APPROVED BY THE HONORS PROGRAM ___________________________________ Dr. Elizabeth Corey, Director DATE: ____________________ ii ORTHOGRAPHIC REFORM AND LANGUAGE PLANNING IN RUSSIAN HISTORY A Thesis Submitted to the Faculty of Baylor University In Partial Fulfillment of the Requirements for the Honors Program By Sylvia Conatser Segura Waco, Texas May 2020 TABLE OF CONTENTS Acknowledgements . iii. Introduction . 1 Chapter One: Peter I and the Rise of Russian Print Culture . 7 Chapter Two: The 20th Century and the Maturation of a Writing System .
    [Show full text]
  • Orthographies in Early Modern Europe
    Orthographies in Early Modern Europe Orthographies in Early Modern Europe Edited by Susan Baddeley Anja Voeste De Gruyter Mouton An electronic version of this book is freely available, thanks to the support of libra- ries working with Knowledge Unlatched. KU is a collaborative initiative designed to make high quality books Open Access. More information about the initiative can be found at www.knowledgeunlatched.org An electronic version of this book is freely available, thanks to the support of libra- ries working with Knowledge Unlatched. KU is a collaborative initiative designed to make high quality books Open Access. More information about the initiative can be found at www.knowledgeunlatched.org ISBN 978-3-11-021808-4 e-ISBN (PDF) 978-3-11-021809-1 e-ISBN (EPUB) 978-3-11-021806-2 ISSN 0179-0986 e-ISSN 0179-3256 ThisISBN work 978-3-11-021808-4 is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License, ase-ISBN of February (PDF) 978-3-11-021809-1 23, 2017. For details go to http://creativecommons.org/licenses/by-nc-nd/3.0/. e-ISBN (EPUB) 978-3-11-021806-2 LibraryISSN 0179-0986 of Congress Cataloging-in-Publication Data Ae-ISSN CIP catalog 0179-3256 record for this book has been applied for at the Library of Congress. ISBN 978-3-11-028812-4 e-ISBNBibliografische 978-3-11-028817-9 Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliogra- fie;This detaillierte work is licensed bibliografische under the DatenCreative sind Commons im Internet Attribution-NonCommercial-NoDerivs über 3.0 License, Libraryhttp://dnb.dnb.deas of February of Congress 23, 2017.abrufbar.
    [Show full text]
  • Downloaded on 2017-02-12T11:34:32Z a Cross-Linguistic Database of Children’S Printed Words in Three Slavic Languages
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Cork Open Research Archive Title A cross linguistic database of children's printed words in three Slavic languages Author(s) Garabik, Radovan; Caravolus, Marketa; Kessler, Brett; Hoeflerova, Eva; Masterson, Jackie; Mikulajova, Marina; Szczerbinski, Marcin; Wierzchon, Piotr Editor(s) Levicka, Jana Garabik, Radovan Publication date 2007 Original citation Garabík, R., Caravolas, M., Kessler, B., Höflerová, E., Masterson, J., Mikulajová, M., Szczerbiński, M., Wierzchoń, P. (2007). 'A cross- linguistic database of children’s printed words in three Slavic languages'. In Levická, J., & Garabík, R. (Eds.). Computer Treatment of Slavic and East European Languages: Fourth International Seminar, Bratislava, Slovakia, 25−27 October 2007: Proceedings (pp. 51−64). Bratislava: Tribun. Type of publication Conference item Link to publisher's http://spell.psychology.wustl.edu/Garabík2007/ version http://korpus.juls.savba.sk/~slovko/2007/ Access to the full text of the published version may require a subscription. Rights © 2007, the respective authors. The articles can be used under the Creative Commons Attribution-ShareAlike 3.0 Unported License. http://creativecommons.org/licenses/by-sa/3.0/ Item downloaded http://hdl.handle.net/10468/736 from Downloaded on 2017-02-12T11:34:32Z A Cross-linguistic Database of Children’s Printed Words in Three Slavic Languages Radovan Garabík1, Markéta Caravolas2, Brett Kessler3, Eva Höflerová4, Jackie Masterson5, Marína Mikulajová6,
    [Show full text]
  • Orthographic and Morphological Correspondences Between Related Slavic Languages As a Base for Modeling of Mutual Intelligibility
    Orthographic and Morphological Correspondences between Related Slavic Languages as a Base for Modeling of Mutual Intelligibility Andrea Fischer, Klára Jágrová, Irina Stenger, Tania Avgustinova, Dietrich Klakow, and Roland Marti Saarland University, Collaborative Research Center (SFB) 1102: Information Density and Linguistic Encoding, Campus A 2.2, 66123 Saarbrücken, Germany Project C4: INCOMSLAV – Mutual Intelligibility and Surprisal in Slavic Intercomprehension {kjagrova, avgustinova}@coli.uni- saarland.de {ira.stenger, rwmslav}@mx.uni- saarland.de {andrea.fischer, dietrich.klakow}@lsv.uni- saarland.de http://www.sfb1102.uni-saarland.de Abstract In an intercomprehension scenario, typically a native speaker of language L1 is confronted with output from an unknown, but related language L2. In this setting, the degree to which the receiver recognizes the unfamiliar words greatly determines communicative success. Despite exhibiting great string-level differences, cognates may be recognized very successfully if the receiver is aware of regular correspondences which allow to transform the unknown word into its familiar form. Modeling L1-L2 intercomprehension then requires the identification of all the regular correspondences between languages L1 and L2. We here present a set of linguistic orthographic correspondences manually compiled from comparative linguistics literature along with a set of statistically-inferred suggestions for correspondence rules. In order to do statistical inference, we followed the Minimum Description Length principle, which proposes to choose those rules which are most effective at describing the data. Our statistical model was able to reproduce most of our linguistic correspondences (88.5% for Czech-Polish and 75.7% for Bulgarian-Russian) and furthermore allowed to easily identify many more non-trivial correspondences which also cover aspects of morphology.
    [Show full text]
  • A Diglossia Approach
    University of Kentucky UKnowledge Linguistics Faculty Publications Linguistics 2002 Slovak Standard Language Development in the 15th–18th Centuries: A Diglossia Approach Mark Richard Lauersdorf University of Kentucky, [email protected] Right click to open a feedback form in a new tab to let us know how this document benefits oy u. Follow this and additional works at: https://uknowledge.uky.edu/lin_facpub Part of the Anthropological Linguistics and Sociolinguistics Commons, Comparative and Historical Linguistics Commons, Other Linguistics Commons, and the Slavic Languages and Societies Commons Repository Citation Lauersdorf, Mark Richard, "Slovak Standard Language Development in the 15th–18th Centuries: A Diglossia Approach" (2002). Linguistics Faculty Publications. 33. https://uknowledge.uky.edu/lin_facpub/33 This Article is brought to you for free and open access by the Linguistics at UKnowledge. It has been accepted for inclusion in Linguistics Faculty Publications by an authorized administrator of UKnowledge. For more information, please contact [email protected]. Slovak Standard Language Development in the 15th–18th Centuries: A Diglossia Approach This article is available at UKnowledge: https://uknowledge.uky.edu/lin_facpub/33 Twenty Mark R. Lauersdorf Slovak Standard Language Development in the 15th–18th Centuries: A Diglossia Approach 1. Introduction Since Charles Ferguson’s initial formal description of “diglossia” in 1959 (Ferguson 1959), research based on this sociolinguistic concept has flourished.1 The framework of diglossia, in its original, Fergusonian configuration and in various expanded and revised forms, has been applied to languages and linguis- tic situations around the world, including many of the Slavic languages.2 Charles Townsend, to whom this Festschrift is dedicated, has himself con- tributed to the body of literature dealing with questions of diglossia in the Slavic-speaking world (e.g., Townsend 1987).
    [Show full text]
  • Orthographic Neighbourhood in Czech Language and Its Use in Research of Reading Impairment Original Study
    Linguistic Frontiers • 3(1) • 54–59 • 2020 DOI: 10.2478/lf-2020-0007 Linguistic Frontiers Orthographic neighbourhood in Czech language and its use in research of reading impairment Original Study Monika Ptáčková, Kateřina Vitásková Palacký University Olomouc Received: June 2020 ; Accepted: June 2020 Abstract : Reading is a complex function that affects our daily living. It requires the cooperation of many structu- res and functions and is influenced by many different factors. One of the factors influencing the word reading process may be the characteristics of the word, such as its frequency of use in the language, its length, ortho- graphic neighbourhood, and others. These characteristics can affect the speed and accuracy of reading in re- aders with and without learning disability. The following paper presents some aspects of our current research focused on the influence of the properties of the word on their reading in the Czech language environment. The concept of orthographic neighbourhood will be explained in more detail including some specifics in its calcu- lation and use in the Czech language. Keywords : reading, orthographic neighbourhood, Damerau-Levenshtein distance, visual word recognition INTRODUCTION Reading is a complex process consisting of many impor- on visual word recognition and reading in people with tant components, such as visual recognition, phonolo- dyslexia (see Davies, 2007) and in the typical population gical and orthographic processing, and others (Grainger (see Grainger 2017). Due to the variance between lan- 2017). There is a variety of different impairments of guages, the results of research studies differ as well1. reading ability, which can be generally divided into two Although in the last ten years many kinds of research in main groups—developmental reading impairments (see various language environments have been performed, e.g.
    [Show full text]
  • Sound Form Signalization in L1 Polish, Czech and Slovak Textbooks
    SOUND FORM SIGNALIZATION IN L1 POLISH, CZECH AND SLOVAK TEXTBOOKS In search of best practices ELŻBIETA AWRAMIUK*, JANA VLČKOVÁ-MEJVALDOVÁ** & ĽUDMILA LIPTÁKOVÁ*** *Uniwersytet w Białymstoku, Poland; **Univerzita Karlova, Czech Republic, *** Prešovská univerzita v Prešove, Slovakia Abstract Phonetic transcription is concerned with how the sounds used in spoken language are represented in written form. In specialized sources, phonetic transcription is a conventionalized notation system; in non- specialist sources, the methods of sound form signalization (SFS) are less conventionalized, but they have important educational functions. The purpose of this study is to present the results of a comparative analysis of several L1 Polish, Czech, and Slovak textbooks to answer the following questions: how sound form is signalized and what practices are best for the development of pupils’ phonetic awareness and more generally for the improvement of their spoken and written communication skills. Textbooks from the second stage of primary schools (Grades 4–6, age 10–13) were analyzed. This quali- tative analysis focuses on searching for instances where orthographic representation changes to fulfil the needs of SFS and where the sound form of language represents the point of didactic interest; it illustrates the function of SFS and its means, as well as compares results obtained in three countries. Keywords: primary school, textbook analysis, respelling, West Slavic languages 1 Awramiuk, E., Vlčková-Mejvaldová, J., Liptáková, Ľ. (2021). Sound form signalization in L1 Polish, Czech and Slovak textbooks: In search of best practices. L1-Educational Studies in Lan- guage and Literature, 21, 1-27. https://doi.org/10.17239/L1ESLL-2021.21.01.01 Corresponding author: Elżbieta Awramiuk, Faculty of Philology, Uniwersytet w Białymstoku, Plac Niezależnego Zrzeszenia Studentów 1, 15-420 Białystok, Poland.
    [Show full text]
  • Historical Perspectives of English Studies in Czech Humanities
    Historical Perspectives of English Studies in Czech Humanities A Working Programme of English Studies for Democritus Association and Linguistic Residualism DEMOCRITUS ASSOCIATION Prague 2001 2 First edition Printed in the Czech Republic 2001 © Pavel Bělíček 2001 ISBN 80-86580-02-4 3 C O N T E N T S HISTORICAL PROGRESS OF LINGUISTIC METHODOLOGY ..... 5 1. The Linguistic Method .................................................. ........................ 5 2. Ancient Origins of Philology .................................................................... 9 3. The Origins of English Philology ............................................................ 13 4. The Systematic Taxonomy of English Linguistic and Literary Studies ... 17 PATHWAYS OF COMPARATIVE LINGUISTICS.............................. 33 1. The Story of Comparative Linguistics ...................................................... 33 2. The Methods of Comparative Linguistics ................................................. 38 3. Applicative Levels in English Studies ...................................................... 41 4. The Growth of the Prague School Methodology....................................... 43 THESES TO MACROLINGUISTIC TYPOLOGY.......................... 52 1. The Typological Approach to Linguistic Change ............................... 52 2. A Typological Approach to English Grammar ........................................ 54 3. A Typological Account of English Phonology ......................................... 60 4. Theses to Macrolinguistics as a Concept of
    [Show full text]
  • Neural Machine Translation Models with Back-Translation for the Extremely Low-Resource Indigenous Language Bribri
    Neural Machine Translation Models with Back-Translation for the Extremely Low-Resource Indigenous Language Bribri Isaac Feldman Dartmouth College [email protected] Rolando Coto-Solano Dartmouth College [email protected] Abstract This paper presents a neural machine translation model and dataset for the Chibchan language Bribri, with an average performance of BLEU 16.9±1.7. This was trained on an extremely small dataset (5923 Bribri-Spanish pairs), providing evidence for the applicability of NMT in extremely low-resource environments. We discuss the challenges entailed in managing training input from languages without standard orthographies, we provide evidence of successful learning of Bribri grammar, and also examine the translations of structures that are infrequent in major Indo-European languages, such as positional verbs, ergative markers, numerical classifiers and complex demonstrative systems. In addition to this, we perform an experiment of augmenting the dataset through iterative back-translation (Sennrich et al., 2016a; Hoang et al., 2018), by using Spanish sentences to create synthetic Bribri sentences. This improves the score by an average of 1.0 BLEU, but only when the new Spanish sentences belong to the same domain as the other Spanish examples. This contributes to the small but growing body of research on Chibchan NLP. 1 Introduction State-of-the-art systems in neural machine translation require large amounts of rich and varied parallel data between the source and target language (Lui et al., 2019). While this is less of a problem for high- resource language pairs like English-German or English-Spanish, there are much fewer resources when one of the languages is an Indigenous language.
    [Show full text]
  • The Language of the Old Czech Legenda "O Svaté Kateřinĕ"
    Slavistische Beiträge ∙ Band 87 (eBook - Digi20-Retro) George M. Cummins The Language of the Old Czech Legenda "O Svaté Kateřinĕ" Verlag Otto Sagner München ∙ Berlin ∙ Washington D.C. Digitalisiert im Rahmen der Kooperation mit dem DFG-Projekt „Digi20“ der Bayerischen Staatsbibliothek, München. OCR-Bearbeitung und Erstellung des eBooks durch den Verlag Otto Sagner: http://verlag.kubon-sagner.de © bei Verlag Otto Sagner. Eine Verwertung oder Weitergabe der Texte und Abbildungen, insbesondere durch Vervielfältigung, ist ohne vorherige schriftliche Genehmigung des Verlages unzulässig. «Verlag Otto Sagner» ist ein Imprint der Kubon & Sagner GmbH. George M. Cummins - 9783954793235 Downloaded from PubFactory at 01/10/2019 05:58:22AM via free access S l a v is t ic h e B eiträge BEGRÜNDET VON ALOIS SCHMAUS HERAUSGEGEBEN VON JOHANNES HOLTHUSEN UND JOSEF SCHRENK REDAKTION: PETER REHDER Band 87 George M. Cummins - 9783954793235 Downloaded from PubFactory at 01/10/2019 05:58:22AM via free access 0004741Б GEORGE М. CUMMINS THE LANGUAGE OF THE OLD CZECH LEGENDA 0 SVATÉ KATERINÈ VERLAG OTTO SAGNER • MÜNCHEN 1975 George M. Cummins - 9783954793235 Downloaded from PubFactory at 01/10/2019 05:58:22AM via free access ן ôayensche bllothek׳6taat8b München J ISBN 3 87690 099 9 Copyright by Verlag Otto Sagner, München 1975 Abteilung der Firma Kubon und Sagner, München Druck: Alexander GroBmann 8 München 19, Ysenburgstraße 7* George M. Cummins - 9783954793235 Downloaded from PubFactory at 01/10/2019 05:58:22AM via free access ACKNOWLEDGMENTS This book is the revised text of my Harvard Ph.D. disser- tation, directed by Horace G. Lunt, and presented to the Depart- ment of Slavic Languages and L iteratures in June, 1973• The following three essays and transcription of the Old Czech Legenda 0 syaté Katefinë do not make up a rigorous textual study in the familiar sense—this work was completed half a century and more ago by the philologists Spina and Gebauer.
    [Show full text]
  • Scripts, Spelling, and Identities
    Scripts, Spelling, and Identities Mark Sebba IN MEMORIAM ALEXANDRA (“MISTY”) JAFFE, 1960–2018 WRITING AS IMAGE: ORTHOGRAPHY AND SCRIPT The written word is pervasive in the contemporary world—probably the most important method of conveying information, conducting social relations, and regulating behavior. But writing is much more than that. Writing, and what we do with it, is fundamentally bound up with our identities. As individuals, we 7 are identified partly by what kinds of literacy we engage in, the kinds of writing we do, how we learn to do it, and the purposes for which it is done. What is true for individuals is also true for larger social groupings, such as nation-states, which are strongly identified with the specific languages—and, in particular, the written languages—they use. While the relevance of written language to identity may be obvious, the role of orthography is less so. In fact, scripts and spellings play a crucial role in identity creation by making choices available at different levels of language structure and different levels of social organization, as this article will show. A well-established principle of sociolinguistics is “where there is choice, there is social meaning.” The choices that people make when they write are part of the process of constructing an identity, both for themselves and for their readers. Unlike spoken language, the written mode always depends on some set of symbols. By definition, there can be no “written language” which does not Mark Sebba works at the Department of Linguistics and English Language, University of Lancaster. His current research focuses on written bilingual and multilingual texts.
    [Show full text]