This thesis suggests a new method of developing reading materials for Navajo students. The core of this method is based on phoneme- grapheme relative frequency correspondence data. A short story was phonetically edited and rewritten using the phoneme-grapheme relative frequency method. This short story was one of the major products of the thesis. This short story and three other short stories previously phonically rewritten, were presented in both this form and in original text to 38 Navajo students for reading. Reading speed and reading comprehension were measured for each student on the short stories. The students performed with superior reading speed and with superior reading comprehension on two stories and equal comprehension on two stories. The difference on reading speed was significant at the .01 level. On the two stories with reading comprehension differences, the difference was significant at the .05 level.

Life would be much easier if there were no typing involved in preparing a document, typing an email, paying online bills, entering credit card details, booking flights, hotels or car rentals online. Imagine a system that would recognize speech and convert it into another form to do these functions automatically. The fact that most people can speak faster than they can type gives a good reason to have a speech recognizer. This thesis concentrates on developing a speaker independent, speech recognizer using Wavelet Packet Transform. Speech corpus in the form of phonemes is collected from an American male and an Indian Female. The subjects chosen for phoneme recognition vary in a number of factors like the accent, gender, age, microphone used to record speech, environment in which phonemes are recorded, etc. These factors increase the complexity of speech recognition. We also assume that the emotions of the speakers are the same and the speakers are stationary while recording phonemes.

Syftet med den här studien var att undersöka hur det särskilda läs-, uttals- och skrivstöd fungerar som elever får i en särskild studioverksamhet på en sfi-skola och vad jag som sfilärare kan tänka på för att stödja eleverna så att de når målen. I samband med detta har jag även studerat och jämfört de fyra elevernas texter i tal och skrift samt deras tankar kring att ha "en bra svenska", självförtroende och om acceptans i det svenska samhället. Studien som är en utvärdering av en särskild stödundervisning för elever på en sfi-skola i Sverige visar med hjälp av tester och intervjuer att verksamheten är ett värdefullt komplement till den ordinarie verksamheten. De fyra intervjuade eleverna menar att man lär sig snabbare i "studion"och får mer individuell tid än i den ordinarie undervisningen. Några fonemtest visar att de fyra eleverna nästan halverat sina fel vid det andra testet, medan fyra andra elever, som aldrig har fått särskilt stöd, endast gör en minskning med nio av totalt 90 fel. Eleverna, som har fått stöd, börjar lära sig de svenska språkljuden i tal och skrift men de gör fortfarande många fel. För de fyra eleverna är det viktigt att lära sig att kommunicera som för dem innebär att kunna prata och göra sig förstådda på svenska. "En bra svenska" menar eleverna betyder att ha ett bra svenskt uttal. Endast den kvinnliga eleven har känt sig diskriminerad eller blivit missförstådd p.g.a. sitt svenska uttal. Enligt resultaten, kan jag och mina kollegor i studion stödja eleverna så att de snabbare lär sig de svenska språkljuden och kan göra sig förstådda i det svenska samhället. / The aim with this study was to examine how the specified reading, pronunciation and writing support functions for students in a special studio at an SFI-school and what I, as an SFIteacher, can do to support students to reach their aims. In connection with this I have also studied and compared the four students' texts in speech and writing as well as their thoughts on having "Good Swedish", confidence and acceptance in the Swedish society. The study which is an evaluation of a special remedial teaching for students at a SFI-school in Sweden shows through tests and interviews that the business is a valuable addition to the regular education. The four students think that they learn faster and get more individual teaching in the "studio" than in their class. Some phoneme tests show that the four students nearly eliminate their errors at the second test, while four other students, who never have had special remedial support, only reduce their errors by nine pieces out of 90. The students who have had remedial support begin to learn Swedish phonemes, speech sounds and spelling, but they still make many errors. For all students it is important to learn to communicate, which means to speak and to be understood in Swedish. "Good Swedish" for the students means to have a good Swedish pronunciation. Only the female student has felt discriminated or misunderstood because of her Swedish pronunciation. According to the results, me and my colleagues in the studio can support the students so that they learn the Swedish language sounds quickly and make themselves understood in Swedish society.

This research expands our understanding of the relationship between orthographic knowledge and phoneme perception by investigating how orthographic knowledge affects phoneme perception not only in the first language (L1) but also in the second language (L2), and an unfamiliar language (L0). Specifically, this research sought not only to confirm that L1 orthographic knowledge influences L1 phoneme perception, but also to determine if L1 orthographic knowledge influences L2 and L0 phoneme perception, particularly as it relates to native English speakers. Via a phoneme counting task, 52 participants were divided into two experimental groups—one with a Russian L0 and one with a Mandarin L0—and counted phonemes in words from their L1 (English) and L0. In addition, two subgroups of participants also counted phonemes in their L2 (either Russian or Mandarin). The stimuli for each language were organized along two parameters: 1) match (half with consistent letter-phoneme correspondences and half with inconsistent correspondences) and 2) homophony (half with cross-language homophonous counterparts and half without homophonous counterparts). The assumption here was that accuracy and RT differences would indicate an effect of orthographic knowledge on phoneme perception. Four-way repeated measures ANOVAs analysed the data along four independent factors: group, language, homophone, and match. Overall, the results support the hypotheses and indicate that L1 orthographic knowledge facilitates L1 and L0 phoneme perception when the words have consistent letter-phoneme correspondences but hinders L1 and L0 phoneme perception when the words have inconsistent correspondences. Similarly, the results indicate that L2 orthographic knowledge facilitates L2 phoneme perception with consistent words but hinders L2 phoneme perception with inconsistent words. On a more specific level, results indicate that not all letter-phoneme mismatches are equal in terms of their effect on phoneme perception, for example mismatches in which one letter represents two sounds (e.g., <x> = /ks/) influence perception more so than do mismatches in which one or more letters are silent (e.g. <sh> = /ʃ/). Findings from this research support previous claims that orthographic and phonological information are co-activated in speech processing even in the absence of visual stimuli (e.g., Blau et al., 2008; Taft et al., 2008; Ziegler & Ferrand 1998), and that listeners are sensitive to orthographic information such that it may trigger unwanted interference when the orthographic and phonological systems provide conflicting information (e.g., Burnham, 2003; Treiman & Cassar, 1997). More importantly, findings show that orthographic effects are not limited to L1. First, phoneme perception in unfamiliar languages (L0) is also influenced by L1 orthography. Second, phoneme perception in L2 is influenced by L2 orthgraphic interference. In fact, L2 orthographic effects appear to override any potential L1 orthographic effects, suggesting orthographic effects are language-specific. Finally, the preliminary findings on the different types of letter-phoneme mismatches show that future research must tease apart the behaviours of different kinds of letter-phoneme inconsistencies. Based on the findings, this dissertation proposes the Bipartite Model of Orthographic Knowledge and Transfer. The model identifies two components within L1 orthographic knowledge: abstract and operational. The model predicts that abstract L1 orthographic knowledge (i.e., the general assumptions and principles about the function of orthography and its relationship to phonology) transfers into nonnative language processing regardless of whether the listeners/speakers are familiar with the nonnatiave language (e.g., Bassetti, 2006; Vokic, 2011). In contrast, the model predicts that operational knowledge (i.e., what letters map to what phonemes) transfers into the nonnative language processing in the absence of nonnative orthographic knowledge (i.e., the L0), but does not transfer in the presence of nonnative orthographic knowledge (i.e., the L2). Rather, L2-specific operational knowledge is created based partly on the transferred abstract knowledge. The research here contributes to the body of literature in four ways. First, the current research supports previous findings and claims regarding orthographic knowledge and native language speech processing. Second, the L2 findings provide insight into the relatively sparse—but growing—understanding of the relationship between L1 and L2 orthography and nonnative speech perception. Third, this research offers a unified (albeit preliminary) account of orthographic knowledge and previous findings by way of the Bipartite Model of Orthographic Knowledge and Transfer. / Graduate

The ability of students to segment the speech stream into phonemic units and to analyze (make judgments as to same or different) beginning consonant phonemes was assessed at grades kindergarten through third from both high and low socioeconomic groups. Segmentation ability was assessed by the use of a test of actual words in a match-to-sample task, a test of synthetic words requiring a same-different judgment and a task which required deletion of a phoneme from a known word to form a new word. Three prerequisite abilities were also assessed: auditory acuity and understanding of the concepts "same" and "different" with regard to sounds, and "beginning" with regard to sequence of sounds.

Password cracking based on dictionary attacks have been confined only to the use of dictionary strings which make sense to both humans and the computer or are usually alphanumeric keyboard patterns. But here we also try to extend the dictionary attacks to homophones which the millennials tend to use more often. The word LOVE is used as LUV, LAV. Based on the pronunciation of a word there can be many spellings to it. Phoneme to Grapheme Correspondences have a great amount of significance here. So here in this research we try to incorporate all such words in the attacking dictionary with the highest possible probabilities to see if it has any impact on the password cracking efficiency. We use the probabilistic context-free grammar password cracker to see what our test results yield.

This dissertation develops a word bank based on a series of well defined filters. The basis of this system is grapheme-phoneme relative frequency control. The word bank is designed to be used in developing reading materials at varied degrees of difficulty. Four short stories were edited as to their grapheme-phoneme relationships. Reading speed and reading comprehension were measured for each student on these stories. All students read a story in the original form and a story in the filtered version. No difference in reading comprehension were found, but reading speed was significantly higher on three of four stories.

Abstract This work was aimed at studying speech perception and auditory performance in the everyday lives of Finnish-speaking postlingually severely or profoundly hearing-impaired adults before and after receiving a multichannel cochlear implant. The association between the formal speech perception results and auditory performance in everyday life was also determined, and an effort was made to define how well a smaller sample represents the nationwide results. The patient series comprised a nationwide retrospective survey (N = 67), in which data on hearing level and word recognition were requested from the hospitals, and a prospective sample from the city of Oulu (N = 20), in whom hearing level, sentence, word and phoneme recognition and phoneme confusions were examined using standardised audiometric measures and formal speech perception tests in a study with a prospective repeated measure design. Categories of auditory performance in everyday life were assessed in both samples. The median sound field hearing level at frequencies of 0.5, 1, 2 and 4 kHz for the subjects in the nationwide survey one year after the switch-on of the implant was comparable to the level of mild hearing impairment. All the subjects achieved at least some open-set word recognition auditorily only (mean 71%, 95% CI 61-81%). The results in the Oulu sample were in line with the nationwide survey. A majority of the subjects (31/40) was able to understand conversation without speechreading one year after switch-on. Sentence recognition by the subjects in the Oulu sample improved most during the initial six months after the switch-on of the implant, whereas word and phoneme recognition improved steadily during the two-year follow-up period. Estimated average sentence recognition after two years was 89% (95% CI 71 to 106%), word recognition 73% (95% CI 58 to 87%), syllable recognition 53% (95% CI 42 to 63%), vowel recognition 80% (95% CI 68 to 92%) and consonant recognition 67% (95% CI 57 to 76%). Confusion of phonemes took place more in the direction a spectral energy distribution at higher frequencies. The association between auditory performance in everyday life and the formal speech perception tests was high (rs > 0.81, p &lt; 0.0001). Systematic prospective assessment of speech perception with tests of differing difficulty is recommended for the follow-up of adult cochlear implant users. / Tiivistelmä Tämän työn tarkoituksena oli tutkia suomenkielisten, kielen oppimisen jälkeen vaikean tai erittäin vaikean kuulovian saaneiden aikuisten kuulon tasoa, puheen vastaanottoa ja kuulon toiminnallista tasoa monikanavaisen sisäkorvaistutteen avulla. Tutkimuksessa selvitettiin myös, miten puheen vastaanottoa mittaavat testit kuvaavat selviytymistä arkipäivän elämässä sisäkorvaistutteen mahdollistaman kuulon avulla. Tarkoituksena oli myös määrittää, millä tavalla pieni otos edustaa kansallisia tuloksia. Tutkimuksessa on retrospektiivinen kansallinen otos (N=67) ja prospektiivinen Oulun otos (N=20). Kansallisessa otoksessa tiedot kuulon tasosta ja sanojen tunnistuskyvystä kerättiin yliopistosairaaloista koehenkilöiden sairauskertomuksista. Oulun otoksessa kuulon tasoa, sekä lauseiden, sanojen ja äänteiden tunnistuskykyä ja äänteiden sekoittuvuuksia tutkittiin audiometrian ja puheenvastaanottoa mittaavien testien avulla kahden vuoden seurannan aikana. Kuulon toiminnallista tasoa arvioitiin kuulon toiminnallisen tason luokituksella molemmissa otoksissa. Kansallisen otoksen koehenkilöiden kuulokynnysten mediaani äänikentässä sisäkorvaistutteella taajuuksilla 0,5, 1, 2 ja 4 kHz oli verrattavissa lievän kuulovian tasoon vuosi sisäkorvaistutteen käyttöönoton jälkeen. Kaikki koehenkilöt kykenivät tunnistamaan vähintään joitain sanoja pelkästään kuulonvaraisesti (keskiarvo 71 %, 95 %:n luottamusväli 61-81 %). Oulun otoksen ja kansallisen otoksen tulokset olivat yhteneväiset. Vuosi sisäkorvaistutteen käyttöönoton jälkeen suurin osa (31/40) koehenkilöistä pystyi keskustelemaan ilman huulioluvun tukea hiljaisessa ympäristössä. Oulun otoksen koehenkilöiden lauseiden tunnistuskyky parani eniten ensimmäisten kuuden kuukauden aikana. Sanojen ja äänteiden tunnistuskyky parani koko kahden vuoden seurannan ajan. Kaksi vuotta sisäkorvaistutteen käyttöönoton jälkeen, estimoitu keskimääräinen lauseiden tunnistusprosentti oli 89 % (95 %:n luottamusväli 71-106 %), sanojen tunnistusprosentti oli 73 % (95 %:n luottamusväli 58-87 %), tavujen tunnistusprosentti oli 53 % (95 %:n luottamusväli 42-63 %), vokaalien tunnistusprosentti oli 80 % (95 %:n luottamusväli 68-92 %) ja konsonanttien tunnistusprosentti oli 67 % (95 %:n luottamusväli 57-76 %). Koehenkilöt sekoittivat vokaaleja ja konsonantteja useimmiten spektraaliselta energialtaan läheisimpään suuremmille taajuuksille sijoittuvaan äänteeseen. Kuulon toiminnallisen tason luokituksen ja puheen vastaanottoa mittaavien testien välinen korrelaatio oli korkea (rs > 0.81, p &lt; 0.0001). Sisäkorvaistutteen saavien aikuisten kuulon tason ja puheen vastaanottokyvyn systemaattinen seuranta vaikeudeltaan eritasoisten testien avulla on tärkeää monipuolisen kuntoutuksen suunnittelun tueksi.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2009. / ENGLISH ABSTRACT: Phoneme recognition systems typically suffer from low classification accuracy. Recognition for South African English is especially difficult, due to the variety of vastly different accent groups. This thesis investigates whether a fusion of classifiers, each trained on a specific accent group, can outperform a single general classifier trained on all. We implemented basic voting and score fusion techniques from which a small increase in classifier accuracy could be seen. To ensure that similarly-valued output scores from different classifiers imply the same opinion, these classifiers need to be calibrated before fusion. The main focus point of this thesis is calibration with the Pool Adjacent Violators algorithm. We achieved impressive gains in accuracy with this method and an in-depth investigation was made into the role of the prior and the connection with the proportion of target to non-target scores. Calibration and fusion using the information metric Cllr was showed to perform impressively with synthetic data, but minor increases in accuracy was found for our phoneme recognition system. The best results for this technique was achieved by calibrating each classifier individually, fusing these calibrated classifiers and then finally calibrating the fused system. Boosting and Bagging classifiers were also briefly investigated as possible phoneme recognisers. Our attempt did not achieve the target accuracy of the classifier trained on all the accent groups. The inherent difficulties typical of phoneme recognition were highlighted. Low per-class accuracies, a large number of classes and an unbalanced speech corpus all had a negative influence on the effectivity of the tested calibration and fusion techniques. / AFRIKAANSE OPSOMMING: Foneemherkenningstelsels het tipies lae klassifikasie akkuraatheid. As gevolg van die verskeidenheid verskillende aksent groepe is herkenning vir Suid-Afrikaanse Engels veral moeilik. Hierdie tesis ondersoek of ’n fusie van klassifiseerders, elk afgerig op ’n spesifieke aksent groep, beter kan doen as ’n enkele klassifiseerder wat op alle groepe afgerig is. Ons het basiese stem- en tellingfusie tegnieke ge¨ımplementeer, wat tot ’n klein verbetering in klassifiseerder akkuraatheid gelei het. Om te verseker dat soortgelyke uittreetellings van verskillende klassifiseerders dieselfde opinie impliseer, moet hierdie klassifiseerders gekalibreer word voor fusie. Die hoof fokuspunt van hierdie tesis is kalibrasie met die Pool Adja- cent Violators algoritme. Indrukwekkende toenames in akkuraatheid is behaal met hierdie metode en ’n in-diepte ondersoek is ingestel oor die rol van die aanneemlikheidswaarskynlikhede en die verwantskap met die verhouding van teiken tot nie-teiken tellings. Kalibrasie en fusie met behulp van die informasie maatstaf Cllr lewer indrukwekkende resultate met sintetiese data, maar slegs klein verbeterings in akkuraatheid is gevind vir ons foneemherkenningstelsel. Die beste resultate vir hierdie tegniek is verkry deur elke klassifiseerder afsonderlik te kalibreer, hierdie gekalibreerde klassifiseerders dan te kombineer en dan die finale gekombineerde stelsel weer te kalibreer. Boosting en Bagging klassifiseerders is ook kortliks ondersoek as moontlike foneem herkenners. Ons poging het nie die akkuraatheid van ons basislyn klassifiseerder (wat op alle data afgerig is) bereik nie. Die inherente probleme wat tipies is tot foneemherkenning is uitgewys. Lae per-klas akkuraatheid, ’n groot hoeveelheid klasse en ’n ongebalanseerde spraak korpus het almal ’n negatiewe invloed op die effektiwiteit van die getoetsde kalibrasie en fusie tegnieke gehad.

Grapheme-based speech recognition systems are faster to develop, but typically do not reach the same level of performance as phoneme-based systems. Using Afrikaans speech recognition as a case study, we first analyse the reasons for the discrepancy in performance, before introducing a technique for improving the performance of standard grapheme-based systems. It is found that by handling a relatively small number of irregular words through phoneme-to-grapheme (P2G) transliteration – transforming the original orthography of irregular words to an ‘idealised’ orthography – grapheme-based accuracy can be improved. An analysis of speech recognition accuracy based on word categories shows that P2G transliteration succeeds in improving certain word categories in which grapheme-based systems typically perform poorly, and that the problematic categories can be identified prior to system development. An evaluation is offered of when category-based P2G transliteration is beneficial and methods to implement the technique in practice are discussed. Comparative results are obtained for a second language (Vietnamese) in order to determine whether the technique can be generalised. / MSc (Computer Science) North-West University, Vaal Triangle Campus, 2014