Speaker diarization : "who spoke when"

Speaker diarization is the process of annotating an input audio with information that attributes temporal regions of the audio signal to their respective sources, which may include both speech and non-speech events. For speech regions, the diarization system also specifies the locations of speaker boundaries and assign relative speaker labels to each homogeneous segment of speech. In short, speaker diarization systems effectively answer the question of ‘who spoke when’. There are several important applications for speaker diarization technology, such as facilitating speaker indexing systems to allow users to directly access the relevant segments of interest within a given audio, and assisting with other downstream processes such as summarizing and parsing. When combined with automatic speech recognition (ASR) systems, the metadata extracted from a speaker diarization system can provide complementary information for ASR transcripts including the location of speaker turns and relative speaker segment labels, making the transcripts more readable. Speaker diarization output can also be used to localize the instances of specific speakers to pool data for model adaptation, which in turn boosts transcription accuracies. Speaker diarization therefore plays an important role as a preliminary step in automatic transcription of audio data. The aim of this work is to improve the usefulness and practicality of speaker diarization technology, through the reduction of diarization error rates. In particular, this research is focused on the segmentation and clustering stages within a diarization system. Although particular emphasis is placed on the broadcast news audio domain and systems developed throughout this work are also trained and tested on broadcast news data, the techniques proposed in this dissertation are also applicable to other domains including telephone conversations and meetings audio. Three main research themes were pursued: heuristic rules for speaker segmentation, modelling uncertainty in speaker model estimates, and modelling uncertainty in eigenvoice speaker modelling. The use of heuristic approaches for the speaker segmentation task was first investigated, with emphasis placed on minimizing missed boundary detections. A set of heuristic rules was proposed, to govern the detection and heuristic selection of candidate speaker segment boundaries. A second pass, using the same heuristic algorithm with a smaller window, was also proposed with the aim of improving detection of boundaries around short speaker segments. Compared to single threshold based methods, the proposed heuristic approach was shown to provide improved segmentation performance, leading to a reduction in the overall diarization error rate. Methods to model the uncertainty in speaker model estimates were developed, to address the difficulties associated with making segmentation and clustering decisions with limited data in the speaker segments. The Bayes factor, derived specifically for multivariate Gaussian speaker modelling, was introduced to account for the uncertainty of the speaker model estimates. The use of the Bayes factor also enabled the incorporation of prior information regarding the audio to aid segmentation and clustering decisions. The idea of modelling uncertainty in speaker model estimates was also extended to the eigenvoice speaker modelling framework for the speaker clustering task. Building on the application of Bayesian approaches to the speaker diarization problem, the proposed approach takes into account the uncertainty associated with the explicit estimation of the speaker factors. The proposed decision criteria, based on Bayesian theory, was shown to generally outperform their non- Bayesian counterparts.

The relationship between hip-abductor strength and the magnitude of pelvic drop in patients with low back pain

Context: It has been theorized that a positive Trendelenburg test (TT) indicates weakness of the stance hip-abductor (HABD) musculature, results in contralateral pelvic drop, and represents impaired load transfer, which may contribute to low back pain. Few studies have tested whether weakness of the HABDs is directly related to the magnitude of pelvic drop (MPD). Objective: To examine the relationship between HABD strength and MPD during the static TT and during walking for patients with nonspecific low back pain (NSLBP) and healthy controls (CON). A secondary purpose was to examine this relationship in NSLBP after a 3-wk HABD-strengthening program. Design: Quasi-experimental. Setting: Clinical research laboratory. Participants: 20 (10 NSLBP and 10 CON). Intervention: HABD strengthening. Main Outcome Measures: Normalized HABD strength, MPD during TT, and maximal pelvic frontal-plane excursion during walking. Results: At baseline, the NSLBP subjects were significantly weaker (31%; P = .03) than CON. No differences in maximal pelvic frontal-plane excursion (P = .72), right MPD (P = 1.00), or left MPD (P = .40) were measured between groups. During the static TT, nonsignificant correlations were found between left HABD strength and right MPD for NSLBP (r = -.32, P = .36) and CON (r = -.24, P = .48) and between right HABD strength and left MPD for NSLBP (r = -.24, P = .50) and CON (r = -.41, P = .22). Nonsignificant correlations were found between HABD strength and maximal pelvic frontal-plane excursion for NSLBP (r = -.04, P = .90) and CON (r = -.14, P = .68). After strengthening, NSLBP demonstrated significant increases in HABD strength (12%; P = .02), 48% reduction in pain, and no differences in MPD during static TT and maximal pelvic frontal-plane excursion compared with baseline. Conclusions: HABD strength was poorly correlated to MPD during the static TT and during walking in CON and NSLBP. The results suggest that HABD strength may not be the only contributing factor in controlling pelvic stability, and the static TT has limited use as a measure of HABD function.

Normative and critical criteria for iliotibial band and iliopsoas muscle flexibility

Context: The Ober and Thomas tests are subjective and involve a "negative" or "positive" assessment, making them difficult to apply within the paradigm of evidence-based medicine. No authors have combined the subjective clinical assessment with an objective measurement for these special tests. Objective: To compare the subjective assessment of iliotibial band and iliopsoas flexibility with the objective measurement of a digital inclinometer, to establish normative values, and to provide an evidence-based critical criterion for determining tissue tightness. Design: Cross-sectional study. Setting: Clinical research laboratory. Patients or Other Participants: Three hundred recreational athletes (125 men, 175 women; 250 in injured group, 50 in control group). Main Outcome Measure(s): Iliotibial band and iliopsoas muscle flexibility were determined subjectively using the modified Ober and Thomas tests, respectively. Using a digital inclinometer, we objectively measured limb position. lnterrater reliability for the subjective assessment was compared between 2 clinicians for a random sample of 100 injured participants, who were classified subjectively as either negative or positive for iliotibial band and iliopsoas tightness. Percentage of agreement indicated interrater reliability for the subjective assessment. Results: For iliotibial band flexibility, the average inclinometer angle was -24.59 degrees +/- 7.27 degrees. A total of 432 limbs were subjectively assessed as negative (-27.13 degrees +/- 5.53 degrees) and 168 as positive (-16.29 degrees +/- 6.87 degrees). For iliopsoas flexibility, the average inclinometer angle was -10.60 degrees +/- 9.61 degrees. A total of 392 limbs were subjectively assessed as negative (-15.51 degrees +/- 5.82 degrees) and 208 as positive (0.34 degrees +/- 7.00 degrees). The critical criteria for iliotibial band and iliopsoas flexibility were determined to be -23.16 degrees and -9.69 degrees, respectively. Between-clinicians agreement was very good, ranging from 95.0% to 97.6% for the Thomas and Ober tests, respectively. Conclusions: Subjective assessments and instrumented measurements were combined to establish normative values and critical criterions for tissue flexibility for the modified Ober and Thomas tests.

The architecture of children’s use of language and tools when problem solving collaboratively with robotics

This paper demonstrates, following Vygotsky, that language and tool use has a critical role in the collaborative problem-solving behaviour of school-age children. It reports original ethnographic classroom research examining the convergence of speech and practical activity in children’s collaborative problem solving with robotics programming tasks. The researchers analysed children’s interactions during a series of problem solving experiments in which Lego Mindstorms toolsets were used by teachers to create robotics design challenges among 24 students in a Year 4 Australian classroom (students aged 8.5–9.5 years). The design challenges were incrementally difficult, beginning with basic programming of straight line movement, and progressing to more complex challenges involving programming of the robots to raise Lego figures from conduit pipes using robots as pulleys with string and recycled materials. Data collection involved micro-genetic analysis of students’ speech interactions with tools, peers, and other experts, teacher interviews, and student focus group data. Coding the repeated patterns in the transcripts, the authors outline the structure of the children’s social speech in joint problem solving, demonstrating the patterns of speech and interaction that play an important role in the socialisation of the school-age child’s practical intellect.

Chromosomal aberrations in squamous cell carcinoma and solar keratoses revealed by comparative genomic hybridization

OBJECTIVE: To identify chromosomal copy numbers of frequent genetic aberrations within squamous cell carcinomas (SCCs) and solar keratoses (SKs), and provide further evidence to support or challenge current dogma concerning the relationship between these lesions. DESIGN: Retrospective analysis of genetic aberrations in DNA from SK and SCC biopsy specimens by comparative genomic hybridization. SETTING: University-based research laboratory in Queensland, Australia. PATIENTS: Twenty-two biopsy specimens from patients with diagnosed SKs (n = 7), cutaneous SCCs (n = 10), or adjoining lesions (n = 5). MAIN OUTCOME MEASURES: Identification of frequent genetic aberrations both specific to SK and SCC and shared by these lesions to investigate their clonal relationship. RESULTS: Shared genomic imbalances were identified in SK and SCC. Frequent gains were located at chromosome arms 3q, 17q, 4p, 14q, Xq, 5p, 9q, 8q, 17p, and 20q, whereas shared regional losses were observed at 9p, 3p, 13q, 17p, 11p, 8q, and 18p. Significant loss of 18q was observed only in SCC lesions. CONCLUSIONS: Our results demonstrate that numerous chromosomal aberrations are shared by the 2 lesions, suggesting a clonal relationship between SK and SCC. Additionally, the genomic loss of 18q may be a significant event in SK progression to SCC. Finally, the type and frequency of aberrations suggests a common mode of tumorigenesis in SCC-derived tumors.

"Listening, hearing and drawing things together to work on solutions" : NAPCAN Play a Part, Queensland, 2010-2013, Evaluation-Research Report

Responding to the idea of child friendly communities, Play a Part is an innovative program advancing preventative strategies for children and young people to minimise exposure to abuse and neglect. The program was developed ensuing an increase in notifications of suspected child abuse and neglect in 2007. Now completing the second phase, the program is a community engagement strategy that aims to prevent child abuse. Play a Part is described as “a whole of community approach to creating child friendly communities” (NAPCAN, 2012). The Play a Part program was piloted between 2007 and 2010 in five southeast Queensland communities, and is currently operating in parts of Logan City region and the Redlands region. To assess the merit of the second phase of the program the Children and Youth Research Centre at Queensland University of Technology was contracted to undertake an evaluation-research at the beginning of 2013.

Parent-teacher engagement : a coteaching and cogenerative dialoguing approach

The accumulated evidence from more than four decades of education research strongly suggests that parent involvement in schools carries significant benefits for students as well as for the success of schools (e.g., Henderson & Mapp, 2002). Governments in Australia and overseas have supported parent involvement in schools with a range of initiatives while parent groups have indicated a strong desire for expanded school roles that include participation in formal educational processes namely curriculum, pedagogy, and assessment. Research has also signalled the need for teachers to engage parents rather than adopt traditional parent-school involvement practices so that parents can participate as joint educators in their children's schooling alongside teachers (Pushor, 2001). Actually improving the quality of contact and relationships between parents and teachers to enable engagement however remains problematic. Coteaching and cogenerative dialoguing originally emerged as an innovative approach in the context of teaching secondary school science. Coteaching brings together the collective expertise of several individuals to expand learning opportunities for students while cogenerative dialogues refer to sessions in which participants talk, listen, and learn from one another about the process (Roth & Tobin, 2002a). Coteaching and cogenerative dialoguing reportedly benefits students academically and socially while rewarding educators professionally and emotionally through the support and collaboration they receive from fellow coteachers. These benefits ensue because coteaching theoretically positions teachers at one another's elbows, providing new and different understandings about teaching based on first-hand perspectives and shared goals for assisting students to learn. This thesis proposes that coteaching and cogenerative dialoguing may provide a vehicle for improving quality of contact and relationships between parents and teachers. To investigate coteaching and cogenerative dialoguing as a parent-teacher engagement mechanism, interpretive ethnographic case study research was conducted involving two parents and a secondary school teacher. Sociological ideas, namely Bourdieu's (1977) fields, habitus, and capitals, together with multiple dialectical concepts such as agency|structure (Sewell, 1992) and agency|passivity (Roth, 2007b, 2010) were assembled into a conceptual framework to examine parent-teacher relationships by describing and explaining cultural production and identity construction throughout the case study. Video and audio recordings of cogenerative dialogues and cotaught lessons comprised the chief data sources. Data were analysed using qualitative techniques such as discourse and conversation analysis to identify patterns and contradictions (Roth & Tobin, 2002a). The use of quality criteria detailed by Guba and Lincoln (2005) gives credence to the way in which ethical considerations infused the planning and conduct of this research. From the processes of data collection and analyses, three broad assertions were proffered. The findings highlight the significance of using multiple coordinated dialectical concepts for analysing the affordances and challenges of coteaching and cogenerative dialogues that include parents and teachers. Adopting the principles and purposes of coteaching and cogenerative dialoguing promoted trusting respectful relationships that generated an equitable culture. The simultaneous processes and tensions between logistics and ethics (i.e., the logistics|ethics dialectic) were proposed as a new way to conceptualise how power was redistributed among the participants. Knowledge of positive emotional energy and ongoing capital exchange conceived dialectically as the reciprocal interaction among cultural, social, and symbolic capitals (i.e., the dialectical relationship of cultural|social|symbolic capital) showed how coteaching and cogenerative dialoguing facilitated mutual understandings, joint decision-making, and group solidarity. The notion of passivity as the dialectical partner of agency explained how traditional roles and responsibilities were reconfigured and individual and collective agency expanded. Complexities that surfaced when implementing the coteaching and cogenerative dialoguing approach were outweighed by the multiple benefits that accrued for all involved. These benefits included the development of community-relevant and culturally-significant curricula that increased student agency and learning outcomes, heightened parent self-efficacy for participating in and contributing to formal educational processes, and enhanced teacher professionalism. This case study contributes to existing theory, knowledge and practice, and methodology in the research areas of parent-teacher relationships, specifically in secondary schools, and coteaching and cogenerative dialoguing. The study is particularly relevant given the challenges schools and teachers increasingly face to meaningfully connect with parents to better meet the needs of educational stakeholders in times of continual, complex, and rapid societal change.

Speaker attribution of Australian broadcast news data

Speaker attribution is the task of annotating a spoken audio archive based on speaker identities. This can be achieved using speaker diarization and speaker linking. In our previous work, we proposed an efficient attribution system, using complete-linkage clustering, for conducting attribution of large sets of two-speaker telephone data. In this paper, we build on our proposed approach to achieve a robust system, applicable to multiple recording domains. To do this, we first extend the diarization module of our system to accommodate multi-speaker (>2) recordings. We achieve this through using a robust cross-likelihood ratio (CLR) threshold stopping criterion for clustering, as opposed to the original stopping criterion of two speakers used for telephone data. We evaluate this baseline diarization module across a dataset of Australian broadcast news recordings, showing a significant lack of diarization accuracy without previous knowledge of the true number of speakers within a recording. We thus propose applying an additional pass of complete-linkage clustering to the diarization module, demonstrating an absolute improvement of 20% in diarization error rate (DER). We then evaluate our proposed multi-domain attribution system across the broadcast news data, demonstrating achievable attribution error rates (AER) as low as 17%.

Assessment of police subjective workload and preference for using a voice-based interface during simulated driving

Police in-vehicle systems include a visual output mobile data terminal (MDT) with manual input via touch screen and keyboard. This study investigated the potential for voice-based input and output modalities for reducing subjective workload of police officers while driving. Nineteen experienced drivers of police vehicles (one female) from New South Wales (NSW) Police completed four simulated urban drives. Three drives included a concurrent secondary task: an imitation licence number search using an emulated MDT. Three different interface output-input modalities were examined: Visual-Manual, Visual-Voice, and Audio-Voice. Following each drive, participants rated their subjective workload using the NASA - Raw Task Load Index and completed questions on acceptability. A questionnaire on interface preferences was completed by participants at the end of their session. Engaging in secondary tasks while driving significantly increased subjective workload. The Visual-Manual interface resulted in higher time demand than either of the voice-based interfaces and greater physical demand than the Audio-Voice interface. The Visual-Voice and Audio-Voice interfaces were rated easier to use and more useful than the Visual-Manual interface, although not significantly different from each other. Findings largely echoed those deriving from the analysis of the objective driving performance data. It is acknowledged that under standard procedures, officers should not drive while performing tasks concurrently with certain invehicle policing systems; however, in practice this sometimes occurs. Taking action now to develop voice-based technology for police in-vehicle systems has potential to realise visions for potentially safer and more efficient vehicle-based police work.

Library Research Support in developing, testing and promoting an open cyberinfrastructure

QUT Library continues to rethink research support with eResearch as a primary driver. The support to the development of the Lens, an open global cyberinfrastructure, has been especially important in the light of technology transfer promotion, and partly in the response to researchers’ needs in following the innovation landscapes not only within the scientific but also patent literature. The Lens http://www.lens.org/lens/ project makes innovation more efficient, fair, transparent and inclusive. It is a joint effort between Cambia http://www.cambia.org.au and Queensland University of Technology (QUT). The Lens serves more than 84 million patent documents in the world as open, annotatable digital public goods that are integrated with scholarly and technical literature along with regulatory and business data. Users can link from search results to visualization and document clusters; from a patent document description to its full-text; from there, if applicable, the sequence data can also be found. Figure 1 shows a BLAST Alignment (DNA) using the Lens. A unique feature of the Lens is the ability to embed search and BLAST results into blogs and websites, and provide real-time updates to them. PatSeq Explorer http://www.lens.org/lens/bio/patseqexplorer allows users to navigate patent sequences that map onto the human genome and in the future, many other genomes. PatSeq Explorer offers three level views for the sequence information and links each group of sequences at the chromosomal level to their corresponding patent documents in the Lens. By integrating sequence and patent search and document clustering capabilities, users can now understand the big and small details on the true extent and scope of genetic sequence patents. QUT Library supported Cambia in developing, testing and promoting the Lens. This poster demonstrates QUT Library’s provision of best practice and holistic research support to a research group and how QUT Librarians have acquired new capabilities to meet the needs of the researchers beyond traditional research support practices.

Research capacity and culture in podiatry : early observations within Queensland Health

Background Research is a major driver of health care improvement and evidence-based practice is becoming the foundation of health care delivery. For health professions to develop within emerging models of health care delivery, it would seem imperative to develop and monitor the research capacity and evidence-based literacy of the health care workforce. This observational paper aims to report the research capacity levels of statewide populations of public-sector podiatrists at two different time points twelve-months apart. Methods The Research Capacity & Culture (RCC) survey was electronically distributed to all Queensland Health (Australia) employed podiatrists in January 2011 (n = 58) and January 2012 (n = 60). The RCC is a validated tool designed to measure indicators of research skill in health professionals. Participants rate skill levels against each individual, team and organisation statement on a 10-point scale (one = lowest, ten = highest). Chi-squared and Mann Whitney U tests were used to determine any differences between the results of the two survey samples. A minimum significance of p < 0.05 was used throughout. Results Thirty-seven (64%) podiatrists responded to the 2011 survey and 33 (55%) the 2012 survey. The 2011 survey respondents reported low skill levels (Median < 4) on most aspects of individual research aspects, except for their ability to locate and critically review research literature (Median > 6). Whereas, most reported their organisation’s skills to perform and support research at much higher levels (Median > 6). The 2012 survey respondents reported significantly higher skill ratings compared to the 2011 survey in individuals’ ability to secure research funding, submit ethics applications, and provide research advice, plus, in their organisation’s skills to support, fund, monitor, mentor and engage universities to partner their research (p < 0.05). Conclusions This study appears to report the research capacity levels of the largest populations of podiatrists published. The 2011 survey findings indicate podiatrists have similarly low research capacity skill levels to those reported in the allied health literature. The 2012 survey, compared to the 2011 survey, suggests podiatrists perceived higher skills and support to initiate research in 2012. This improvement coincided with the implementation of research capacity building strategies.

What makes school leaders inspirational and how does this relate to mentoring?

Leadership comes in many forms (such as transactional, transformational, distributed) and its effectiveness can inspire others to achieve organisational goals and visions. Inspiration as an emotional event requires receptiveness and an awareness of social interdependence. When mentees are inspired by mentor role models they can extend personal attributes and practices. Similar to other leaders, inspiring mentors can motivate mentees to develop a strength of character and achieve goals in the workplace. What makes school leaders inspirational and how does this relate to mentoring? This qualitative study collects data from 25 experienced teachers, which involved written questionnaire, work samples, and audio-recorded focus group discussions. These participants indicated that inspirational school leaders were those who had: (1) organisational goals (e.g., visionary, goal driven, innovative, & motivational); (2) professional skills such as being knowledgeable, communicative, and acknowledging others’ achievements; and (3) personal attributes (e.g., integrity, active listening, respectful, enthusiastic, & approachable). This research shows how mentors and school leaders can consider the inspirational attributes and practices outlined by participants in this study to inspire teaching staff. For example, an awareness of attentive listening, motivational and visionary practices, and acknowledging individual achievements can guide school leaders and mentors to inspire others for achieving organsational goals and visions.

Desirable attributes and practices for mentees : mentor teachers' expectations

Research indicates attributes and practices for mentor teachers that can be used for effective mentoring. Universities provide guidelines for preservice teacher (mentee) engagement in schools generally from anecdotal evidence, however, what are desirable attributes and practices for mentees? This qualitative study gathers data from 25 mentor teachers through extended response questionnaire and audio-recorded focus group discussions about attributes and practices for mentees. Findings showed that desirable attributes for mentees included: enthusiasm, being personable, commitment to children, lifelong learning/love of learning, open/reflective to feedback, develop resilience, and taking responsibility for their learning, while desirable practices included: planned and preparation for teaching, reflective practices, understanding school and university policies, knowing students for differentiated learning, and building a teaching repertoire (e.g. teaching strategies, behaviour management, content knowledge, and questioning skills). Preservice teachers need to consider teachers’ suggestions on desirable attributes and practices that can help them achieve positive teaching experiences.

Translating vision research into policy and practice to improve the visibility, and hence safety, or road workers at night

The process of translating research into policy and practice is not well understood. This paper uses a case study approach to interpret an example of translation with respect to theoretical approaches identified in the literature. The case study concerns research into “biological motion” or “biomotion”: when lights are placed on the moveable joints of the body and the person moves in a dark setting, there is immediate and accurate recognition of the human form although only the lights can be seen. QUT was successful in gaining Australian Research Council funding with the support of the predecessors of the Queensland Department of Transport and Main Roads (TMR) to research the biomotion effect in road worker clothing using reflective tape rather than lights, and this resulted in the incorporation of biomotion marking into AS/NZS 4602.1 2011. The most promising approach to understanding the success of this translation, SWOV’s “knowledge utilisation approach” provided some insights but was more descriptive than predictive and provided “necessary but not sufficient” conditions for translation. In particular, the supportive efforts of TMR staff engaged in the review and promulgation of national standards were critical in this case. A model of the conclusions is presented. The experiences gained in this case should provide insights into the processes involved in effectively translating research into practice.

Rapid yet accurate speech indexing using dynamic match lattice spotting

The support for typically out-of-vocabulary query terms such as names, acronyms, and foreign words is an important requirement of many speech indexing applications. However, to date many unrestricted vocabulary indexing systems have struggled to provide a balance between good detection rate and fast query speeds. This paper presents a fast and accurate unrestricted vocabulary speech indexing technique named Dynamic Match Lattice Spotting (DMLS). The proposed method augments the conventional lattice spotting technique with dynamic sequence matching, together with a number of other novel algorithmic enhancements, to obtain a system that is capable of searching hours of speech in seconds while maintaining excellent detection performance