The aim of this study was to develop, digitally record, evaluate, and psychometrically equate a set of Taiwan Mandarin bisyllabic word lists to be used for word recognition testing. Frequently used bisyllabic words were selected and digitally recorded by male and female talkers of Taiwan Mandarin. Twenty normally hearing subjects were presented each word to find the percentage of words which they could correctly recognize. Each word was measured at 10 intensity levels (-5 to 40 dB HL) in increments of 5 dB. Logistic regression was used to include 200 words with the steepest logistic regression slopes in four psychometrically equivalent word lists of 50 words each with eight half-lists of 25 words each. Digital recordings of the psychometrically equivalent bisyllabic word recognition lists are available on compact disc.

The purpose of this study was to develop digitally recorded speech audiometry materials in the Japanese language to evaluate Speech Reception Threshold (SRT) and speech discrimination. Trisyllabic words were used to evaluate the SRT and bisyllabic words were used for speech discrimination. Words were recorded by one native female talker and one native male talker who were judged as having standard Japanese dialects. Twenty native Japanese speakers between the ages of 20 and 32 were used as subjects to evaluate 69 trisyllabic words across 13 different intensity levels. The 25 trisyllabic words with the steepest psychometric function (%/dB) were selected for inclusion in the final CD. The final trisyllabic words were digitally adjusted so that the threshold of each word was equal to the mean PTA (3.42 dB HL) of all the subjects. The mean psychometric function (%/dB) at 50% for the trisyllabic words was 9.6 %/dB for the male talker and 7.7 %/dB for the female talker. The same 20 subjects were also used to evaluate 240 bisyllabic words across 10 different intensity levels. A logistic regression was used to obtain regression slopes for each of the 240 words. The 200 bisyllabic words with the steepest slope were selected for inclusion in the final CD. Four lists of 50 words each and eight half-lists of 25 words each were created from the selected bisyllabic words. A chi-square statistic revealed no significant differences among the lists or half-lists. The mean psychometric function at 50% for the bisyllabic lists and half-lists was 5.9 %/dB for the male talker and 5.2 %/dB for the female talker.

The purpose of this study was to create psychometrically equivalent word lists in the language of Cantonese for word recognition testing. Frequently used bisyllabic Cantonese words were recorded by a native female and male talker. The word lists were evaluated by administering the word recognition lists to 20 native speakers of Cantonese with normal hearing. Each list was presented at 10 different intensity levels ranging from -5 to 40 dB HL in 5 dB increments. Logistic regression was used to determine the words with the steepest logistic regression slopes. The 200 words with the steepest slopes were then formulated into four lists of 50 words and eight half-lists of 25 words. The mean psychometric slope value at the 50% location for the male talker was 7.5%/dB while the mean slope for the female talker was slightly steeper at 7.6%/dB. The word lists were digitally recorded on compact discs for worldwide use.

The purpose of this study was to evaluate an updated list of digitally recorded Speech Recognition Threshold (SRT) materials for test-retest reliability. Chipman (2003) identified 33 psychometrically equated spondaic words that are frequently occurring in English today. These digitally recorded words were used to determine the SRT of 40 participants using the American Speech-Language Hearing Association guidelines. The participants were between the ages of 19 and 83 years and presented with hearing impairment ranging from normal to severe. The individual's pure-tone averages classified 16 participants with normal hearing to slight loss, 12 participants with mild loss, and 12 participants with moderate to severe hearing loss. The speech materials were presented to participants in one randomly selected ear. The SRT was measured for the same ear in both the test and retest conditions. The average SRT for the test condition was 22.7 dB HL and 22.8 dB HL in the retest condition with an improvement of 0.1 dB for retest but no significant difference was identified. Using a modified variance equation to determine test-retest reliability resulted in a 0.98, indicating almost perfect reliability. Therefore the test-retest reliability was determined to be exceptional for the new SRT words.

We propose the fuzzy lexical representations (FLRs) hypothesis that regards fuzziness as a core property of nonnative (L2) lexical representations (LRs). Fuzziness refers to imprecise encoding at different levels of LRs and interacts with input frequency during lexical processing and learning in adult L2 speakers. The FLR hypothesis primarily focuses on the encoding of spoken L2 words. We discuss the causes of fuzzy encoding of phonological form and meaning as well as fuzzy form-meaning mappings and the consequences of fuzzy encoding for word storage and retrieval. A central factor contributing to the fuzziness of L2 LRs is the fact that the L2 lexicon is acquired when the L1 lexicon is already in place. There are two immediate consequences of such sequential learning. First, L2 phonological categorization difficulties lead to fuzzy phonological form encoding. Second, the acquisition of L2 word forms subsequently to their meanings, which had already been acquired together with the L1 word forms, leads to weak L2 form-meaning mappings. The FLR hypothesis accounts for a range of phenomena observed in L2 lexical processing, including lexical confusions, slow lexical access, retrieval of incorrect lexical entries, weak lexical competition, reliance on sublexical rather than lexical heuristics in word recognition, the precedence of word form over meaning, and the prominence of detailed, even if imprecisely encoded, information about LRs in episodic memory. The main claim of the FLR hypothesis – that the quality of lexical encoding is a product of a complex interplay between fuzziness and input frequency – can contribute to increasing the efficiency of the existing models of LRs and lexical access.

No description available.

No description available.

<p>Neural responses to anticipatory coarticulatory cues were investigated across systematically varying phonological conditions. Congruent or incongruent subphonemic information was placed between an initial consonant and a vowel in a consonant-vowel- consonant (CVC) spoken word (Archibald & Joanisse, 2011). Due to physical and temporal differences across sound classes, the objective was to investigate whether coarticulatory information would be processed differently across controlled manipulations of onset (fricative vs. stop) and vowel type (height vs. backness). Event- related potentials (ERPs) were recorded during a printed-word/spoken-word matching paradigm, in which participants indicated whether a visual prime stimulus and a spoken word matched/mismatched. The “Phonological Mapping Negativity” (PMN) component provides strong evidence that the use of coarticulatory information in speech recognition varies in strength and timing as a function of onset type (fricative vs. stop) and vowel height (high vs. low). Coarticulatory cues were more readily perceived in spoken word beginning with fricatives than with stops. Similarly, subphonemic variations were more easily detected in low vowels than in high vowels. Observed perceptual and temporal differences are interpreted to reflect variations in subphonemic and phonological processing.</p> / Master of Science (MSc)

In this thesis, we address some of the challenges involved in developing a robust writer-independent, lexicon-free system to recognize online Tamil words. Tamil, being a Dravidian language, is morphologically rich and also agglutinative and thus does not have a finite lexicon. For example, a single verb root can easily lead to hundreds of words after morphological changes and agglutination. Further, adoption of a lexicon-free recognition approach can be applied to form-filling applications, wherein the lexicon can become cumbersome (if not impossible) to capture all possible names. Under such circumstances, one must necessarily explore the possibility of segmenting a Tamil word to its individual symbols. Modern day Tamil alphabet comprises 23 consonants and 11 vowels forming a total combination of 313 characters/aksharas. A minimal set of 155 distinct symbols have been derived to recognize these characters. A corpus of isolated Tamil symbols (IWFHR database) is used for deriving the various statistics proposed in this work. To address the challenges of segmentation and recognition (the primary focus of the thesis), Tamil words are collected using a custom application running on a tablet PC. A set of 10000 words (comprising 53246 symbols) have been collected from high school students and used for the experiments in this thesis. We refer to this database as the ‘MILE word database’. In the first part of the work, a feedback based word segmentation mechanism has been proposed. Initially, the Tamil word is segmented based on a bounding box overlap criterion. This dominant overlap criterion segmentation (DOCS) generates a set of candidate stroke groups. Thereafter, attention is paid to certain attributes from the resulting stroke groups for detecting any possible splits or under-segmentations. By relying on feedbacks provided by a priori knowledge of attributes such as number of dominant points and inter-stroke displacements the recognition label and likelihood of the primary SVM classifier linguistic knowledge on the detected stroke groups, a decision is taken to correct it or not. Accordingly, we call the proposed segmentation as ‘attention feedback segmentation’ (AFS). Across the words in the MILE word database, a segmentation rate of 99.7% is achieved at symbol level with AFS. The high segmentation rate (with feedback) in turn improves the symbol recognition rate of the primary SVM classifier from 83.9% (with DOCS alone) to 88.4%. For addressing the problem of segmentation, the SVM classifier fed with the x-y trace of the normalized and resampled online stroke groups is quite effective. However, the performance of the classifier is not robust to effectively distinguish between many sets of similar looking symbols. In order to improve the symbol recognition performance, we explore two approaches, namely reevaluation strategies and language models. The reevaluation techniques, in particular, resolve the ambiguities in base consonants, pure consonants and vowel modifiers to a considerable extent. For the frequently confused sets (derived from the confusion matrix), a dynamic time warping (DTW) approach is proposed to automatically extract their discriminative regions. Dedicated to each confusion set, novel localized cues are derived from the discriminative region for their disambiguation. The proposed features are quite promising in improving the symbol recognition performance of the confusion sets. Comparative experimental analysis of these features with x-y coordinates are performed for judging their discriminative power. The resolving of confusions is accomplished with expert networks, comprising discriminative region extractor, feature extractor and SVM. The proposed techniques improve the symbol recognition rate by 3.5% (from 88.4% to 91.9%) on the MILE word database over the primary SVM classifier. In the final part of the thesis, we integrate linguistic knowledge (derived from a text corpus) in the primary recognition system. The biclass, bigram and unigram language models at symbol level are compared in terms of recognition performance. Amongst the three models, the bigram model is shown to give the highest recognition accuracy. A class reduction approach for recognition is adopted by incorporating the language bigram model at the akshara level. Lastly, a judicious combination of reevaluation techniques with language models is proposed in this work. Overall, an improvement of up to 4.7% (from 88.4% to 93.1%) in symbol level accuracy is achieved. The writer-independent and lexicon-free segmentation-recognition approach developed in this thesis for online handwritten Tamil word recognition is promising. The best performance of 93.1% (achieved at symbol level) is comparable to the highest reported accuracy in the literature for Tamil symbols. However, the latter one is on a database of isolated symbols (IWFHR competition test dataset), whereas our accuracy is on a database of 10000 words and thus, a product of segmentation and classifier accuracies. The recognition performance obtained may be enhanced further by experimenting on and choosing the best set of features and classifiers. Also, the word recognition performance can be very significantly improved by using a lexicon. However, these are not the issues addressed by the thesis. We hope that the lexicon-free experiments reported in this work will serve as a benchmark for future efforts.

Arising from the increasing demands of the twenty-first century workplace, concern over learner reading performance is at the forefront of national education. The increasing demands have raised the literacy bar for learners and subsequently, schools have been forced to accommodate instruction for these increased expectations. Successful reading requires the learner to incorporate a number of reading skills in appropriate ways. Oral reading fluency and reading comprehension are identified as components in effectively gaining meaning from text. A reciprocal relationship exist between the two that allows one to comprehend more thoroughly as one reads more fluently. Additionally, as one reads more fluently, one‟s ability to comprehend also improves. This is due to the fact that one‟s brain is more capable in processing text when one is able to read fluently. Therefore, when one automatically identifies words one is able to comprehend text more completely. The purpose of this study was to determine whether a linear relationship exists between Grade 4 learners‟ oral reading fluency on different types of tests and their reading comprehension. The study was conducted within a positivistic research paradigm. A one-shot crosssectional survey design was used to determine the relationship between oral reading fluency and reading comprehension of Grade 4 learners in selected schools in Kimberley in the Northern Cape Province. Five schools, representing the different quintiles, were selected to participate in the study. A total of 406 Grade 4 learners made up the study population. Two tests were developed and validated in order to assess the learners‟ oral reading fluency and reading comprehension. The data was analysed by means of descriptive statistics as well as Pearson product-moment correlations. The results indicate that learners in rural schools could only read at 52 words per minute (wpm) which meant that they could be grouped in the 10th percentile. With regard to reading comprehension the learners in the rural areas scored an average of 54% on the first reading comprehension test. Their results on the second comprehension test indicated that they experienced difficulties with inference questions. The results indicated that learners in urban schools read at 107.5 words per minute (wpm) which meant that they could be grouped between the 50th and 75th percentile. In their first reading comprehension test they scored an average of 78%. Their results on the second comprehension test indicated that they experienced difficulties with interpretation questions. Pearson product moment correlations indicated a practically significant difference between rural and urban schools on oral reading fluency and reading comprehension with urban schools outperforming rural schools. Overall, the results indicated a practically significant relationship of r = 0.69 between oral reading fluency and reading comprehension. The findings of this study should be noted by teachers as well as be addressed in interventions as a matter of urgency. / MEd (Learner Support), North-West University, Potchefstroom Campus, 2014