Towards improved speech recognition for resource poor languages

In recent times, the improved levels of accuracy obtained by Automatic Speech Recognition (ASR) technology has made it viable for use in a number of commercial products. Unfortunately, these types of applications are limited to only a few of the world’s languages, primarily because ASR development is reliant on the availability of large amounts of language specific resources. This motivates the need for techniques which reduce this language-specific, resource dependency. Ideally, these approaches should generalise across languages, thereby providing scope for rapid creation of ASR capabilities for resource poor languages. Cross Lingual ASR emerges as a means for addressing this need. Underpinning this approach is the observation that sound production is largely influenced by the physiological construction of the vocal tract, and accordingly, is human, and not language specific. As a result, a common inventory of sounds exists across languages; a property which is exploitable, as sounds from a resource poor, target language can be recognised using models trained on resource rich, source languages. One of the initial impediments to the commercial uptake of ASR technology was its fragility in more challenging environments, such as conversational telephone speech. Subsequent improvements in these environments has gained consumer confidence. Pragmatically, if cross lingual techniques are to considered a viable alternative when resources are limited, they need to perform under the same types of conditions. Accordingly, this thesis evaluates cross lingual techniques using two speech environments; clean read speech and conversational telephone speech. Languages used in evaluations are German, Mandarin, Japanese and Spanish. Results highlight that previously proposed approaches provide respectable results for simpler environments such as read speech, but degrade significantly when in the more taxing conversational environment. Two separate approaches for addressing this degradation are proposed. The first is based on deriving better target language lexical representation, in terms of the source language model set. The second, and ultimately more successful approach, focuses on improving the classification accuracy of context-dependent (CD) models, by catering for the adverse influence of languages specific phonotactic properties. Whilst the primary research goal in this thesis is directed towards improving cross lingual techniques, the catalyst for investigating its use was based on expressed interest from several organisations for an Indonesian ASR capability. In Indonesia alone, there are over 200 million speakers of some Malay variant, provides further impetus and commercial justification for speech related research on this language. Unfortunately, at the beginning of the candidature, limited research had been conducted on the Indonesian language in the field of speech science, and virtually no resources existed. This thesis details the investigative and development work dedicated towards obtaining an ASR system with a 10000 word recognition vocabulary for the Indonesian language.

Mastering medical terminology : Australia and New Zealand

Mastering Medical Terminology: Australia and New Zealand is medical terminology book of relevance to an audience in Australia and New Zealand. Australian terminology, perspectives, examples and spelling have been included and Australian pronunciation specified. The textbook is accompanied by a self-help workbook, an online workbook and a Smartphone app. Throughout Mastering Medical Terminology, review of medical terminology as it is used in clinical practice is highlighted. Features of the textbook, workbook and electronic product include: • Simple, non-technical explanations of medical terms • Workbook format with ample spaces to write answers • Explanations of clinical procedures, laboratory tests and abbreviations used in Australian clinical practice, as they apply to each body system and speciality area • Pronunciation of terms and spaces to write meanings of terms • Practical applications sections • Exercises that test understanding of terminology as students work through the text chapter by chapter • Review activities that pull together terminology to help students study • Comprehensive glossary and appendices for reference • Links to other useful references, such as websites and textbooks.

Australian Aboriginal words in English : their origin and meaning [Second edition]

Australian Aboriginal Words in English records the Aboriginal contribution to Australian English and provides a fascinating insight into the contact between the first Australians and European settlers. The words are grouped according to subject, and for each one there is information on the Aboriginal language from which it derives, the date of its first written use in English, and its present meaning and pronunciation. This book brings them together and provides the fullest available information about their Aboriginal background and their Australian English History.

MedWords : Australia and New Zealand

MedWords is the essential accompaniment to the Mastering Medical Terminology suite of learning tools.Learn correct pronunciation by listening to expert Australian voice recordings of over 2,000 medical terms. Practice by recording your own voice

Èvoljucija sistemy glasnyh fonem v nekotoryh russkih govorah Vologodskoj oblasti

The present dissertation analyses 36 local vernaculars of villages surrounding the northern Russian city of Vologda in relation to the system of the vowels in the stressed syllables and those preceding the stressed syllables by using the available dialectological researches. The system in question differs from the corresponding standard Russian system by that the palatalisation of the surrounding consonants affects the vowels much more significantly in the vernaculars, whereas the phonetic difference between the stressed and non-stressed vowels is less obvious in them. The detailed information on the local vernaculars is retrieved from the Dialektologičeskij Atlas Russkogo Jazyka dialect atlas, the data for which were collected, for the most part, in the 1940 s and 1950 s. The theoretical framework of the research consists of a brief cross-section of western sociolinguistic theory related to language change and that of historical linguistics related to the Slavonic vowel development, which includes some new theories concerning the development of the Russian vowel phonemes. The author has collected dialect data in one of the 36 villages and three villages surrounding it. During the fieldwork, speech of nine elderly persons and ten school children was recorded. The speech data were then transcribed with coded information on the corresponding etymological vowels, the phonetic position, and the factual pronunciation at each appearance of vowels in the phonetic positions named above. The data from both of the dialect strata were then systematised to two corresponding systems that were compared with the information retrievable from the dialect atlas and other dialectological literature on the vowel phoneme system of the traditional local vernacular. As a result, it was found out (as hypothesised) that the vernacular vowel phoneme system has approached that of the standard language but has nonetheless not become similar to it. The phoneme quantity of the traditional vernacular is by one greater than that of the standard language, whereas the vowel phoneme quantity in the speech of the school children coincides with that in the standard language, although the phonetic realisations differ to some extent. The analysis of the speech of the elderly people resulted in that it is quite difficult to define the exact phoneme quantity of this stratum due to the fluctuation and irregularities in the realisation of the old phoneme that has ceased to exist in the newest stratum. It was noticed that the effect of the quality of the surrounding consonants on the phonetic realisation of the vowel phonemes has diminished, and the dependence of the phonetic realisation of a vowel phoneme on its place in a word in relation to the word stress has become more and more obvious, which is the state of affairs in the standard language as well.

"Mut ei tää oo hei midsommarista!" - ruotsin kielen viestinnällinen suullinen harjoittelu yhteistoiminnallisten skeema- ja elaborointitehtävien avulla

Communicative oral practice in Swedish through collaborative schema-based and elaboration tasks The general aim of this study was to learn how to better understand foreign language communicative oral practice and to develop it as part of communicative language teaching. The language-specific aim was to study how Swedish was being practised communicatively and orally in a classroom context as part of the didactic teaching-studying-learning process, and how the students' communicative oral practice in Swedish was carried out through collaborative schema-based and elaboration tasks. The scientific problem of this study focused on the essence of foreign language communicative oral proficiency. The research questions were concerned with 1) the students' involvement in carrying out the given oral tasks; 2) the features of communication and interaction strategies; 3) thematic vocabulary, and 4) the students' experiences and conceptions of the communicative oral tasks used. The study consisted of two groups of students from a Helsinki-area school (a group of upper secondary school students, Swedish Level A, Courses 2 and 3, n=9; and a group of basic education students, Swedish Level B, Course 2, n=13). The study was carried out as a pedagogically oriented case study which included certain features of ethnographic research and where the students' teacher acted as a researcher of her own work. The communicative oral practice contained five different tasks. The research data were gathered through systematic observation, audio recordings and by a questionnaire. The data were analysed through ethnographic content analysis methods. The main research finding was that a good deal of social interaction, collaboration and communication took place between the students when involved in communicative oral practice in Swedish. The students took almost optimal advantage of the allocated training time. They mostly used Swedish when participating in interactional communication. Finnish was mostly used by the students when they were deciding how to carry out a given task, aiming at intersubjectivity or negotiating meaning. The students were relaxed when practising Swedish. They also asked for and gave linguistic help in the spirit of collaborative learning principles. This resulted in interaction between students that highlighted certain features of negotiation of meaning, scaffolding and collaborative dialogue. Asking for and giving help in language issues concentrated mainly on vocabulary, and only in a few cases on grammar or pronunciation. The students also needed the teacher as a mentor. As well, the students had an enjoyable time when practising, which was most often related to carrying out the oral tasks. The thematic vocabulary used by the students corresponded well to the thematic lexis that served as a basis for the practice. At its most efficient, this lexis was most evident when the basic education students were carrying out schema-based tasks. The students' questionnaire answers agreed with the research findings gained through systematic observation and the analysis of audio recordings. The communicative tasks planned by the teacher and implemented by the students were very much in line. The language-didactic theory as presented in this study and the research findings can be widely utilised in pre-service and in-service teacher education, as well as, more generally, when developing communicative language teaching. Key words: communicative oral practice; the Swedish language; foreign language; didactic teaching-studying-learning process; communicative language teaching; collaborative task; schema-based task; elaboration task.

Mig eller mej, själ eller sjel? Problem och lösningar vid transkription av svenska sångtexter

Abstract (Mig or mej, själ or sjel? Problems and solutions in the transcription of Swedish song texts): In this article I am pointing out and discussing problems and solutions concerning phonetic transcription of Swedish song texts. My material consists of 66 Swedish songs phonetically transcribed. The transcriptions were published by The Academy of Finnish Art Song in 2009. The first issue was which level of accuracy should be chosen. The transcriptions were created to be clear at a glance and suitable for the needs of interpretation of non Swedish speaking singers. The principle was to use as few signs and symbols as possible without sacrificing accuracy. Certain songs were provided with additional information whenever there was a chance of misinterpretation. The second issue was which geographic variety of the language should be visible in the transcription, Standard Swedish or Finland-Swedish? The songs in the volume are a selection of well-known works that are also of international interest. Most were composed by Jean Sibelius (1865–1957), a substantial number of whose songs were based on poems written by Finland’s national poet, Johan Ludvig Runeberg (1804–1877). Thus I chose to use the variety of Swedish language spoken in Finland, in order to reflect the cultural origin of the songs. This variety differs slightly from the variety spoken in Sweden both on prosodic and phonetic level. In singing, the note-text gives the interpretor enough information about prosody. The differences concern mostly the phonemes. A fully consequent transcript was, however, difficult to make, due to vocal requirement. So, for example, in an unstressed final syllable the vowel was often indicated as a central vowel, which in singing is given a more direct emphasis than in a literal pronunciation, even if this central vowel does not occur in spoken Finland-Swedish.

Vieraan aksentin arviointi ja mittaaminen suomessa

In the field of second language (L2) acquisition, the term `foreign accent´ is often used to refer to speech characteristics that differ from the pronunciation of native speakers. Foreign accent may affect the intelligibility and perceived comprehensibility of speech and it is also sometimes associated with negative attitudes. The degree of L2 learners foreign accent and the speech characteristics that account for it have previously been studied through speech perception experiments and acoustic measurements. Perception experiments have shown that native listeners are easily able to identify foreign accent in speech. However to date, no studies have been done on the assessment of foreign accent in the speech of non-native speakers of Finnish. The aim of this study is to examine how native speakers of Finnish rate the degree of foreign accentedness in the speech of Russian L2 learners of Finnish. Furthermore, phonetic analysis is used to study the characteristics of speech that affect the perceived strength of foreign accent. Altogether 96 native speakers of Finnish listened to excerpts of read-aloud and spontaneous Finnish speech from ten Russian and six Finnish female speakers. The Russian speakers were intermediate and advanced learners of Finnish and had all immigrated to Finland as adults. Among the listeners, was a group of teachers of Finnish as an L2, and it was presumed that these teachers had been exposed to foreign accent in Finnish and were used to hearing it. The temporal aspects and segmental properties of speech were phonetically analysed in the speech of the Russian speakers in order to measure their effect on the perceived degree of accent. Although wide differences were observed in the use of the rating scale among the listeners, they were still quite unanimous on which speakers had the strongest foreign accent and which had the mildest. The listeners background factors had little effect on their ratings, and the ratings of the teachers of Finnish as an L2 did not differ from those of the other listeners. However, a clear difference was noted in the ratings of the two types of stimuli used in the perception experiment: the read-aloud speech was rated as more strongly accented than the spontaneous speech. It is important to note that the assessment of foreign accent is affected by many factors and their complex interactions in the experimental setting. Futher the study found that, both the temporal aspects of speech, often associated with fluency, and the number of single deviant phonetic segments contributed to the perceived degree of accentedness in the speech of the native Russian speakers.

Development of the 2003 CU-HTK conversational telephone speech transcription system

This paper describes the development of the 2003 CU-HTK large vocabulary speech recognition system for Conversational Telephone Speech (CTS). The system was designed based on a multi-pass, multi-branch structure where the output of all branches is combined using system combination. A number of advanced modelling techniques such as Speaker Adaptive Training, Heteroscedastic Linear Discriminant Analysis, Minimum Phone Error estimation and specially constructed Single Pronunciation dictionaries were employed. The effectiveness of each of these techniques and their potential contribution to the result of system combination was evaluated in the framework of a state-of-the-art LVCSR system with sophisticated adaptation. The final 2003 CU-HTK CTS system constructed from some of these models is described and its performance on the DARPA/NIST 2003 Rich Transcription (RT-03) evaluation test set is discussed.

Automatic transcription of conversational telephone speech

This paper discusses the Cambridge University HTK (CU-HTK) system for the automatic transcription of conversational telephone speech. A detailed discussion of the most important techniques in front-end processing, acoustic modeling and model training, language and pronunciation modeling are presented. These include the use of conversation side based cepstral normalization, vocal tract length normalization, heteroscedastic linear discriminant analysis for feature projection, minimum phone error training and speaker adaptive training, lattice-based model adaptation, confusion network based decoding and confidence score estimation, pronunciation selection, language model interpolation, and class based language models. The transcription system developed for participation in the 2002 NIST Rich Transcription evaluations of English conversational telephone speech data is presented in detail. In this evaluation the CU-HTK system gave an overall word error rate of 23.9%, which was the best performance by a statistically significant margin. Further details on the derivation of faster systems with moderate performance degradation are discussed in the context of the 2002 CU-HTK 10 × RT conversational speech transcription system. © 2005 IEEE.