Speaker diarization: Segmentation and clustering of speeches

Speaker diarization is the process of sorting speeches according to the speaker. Diarization helps to search and retrieve what a certain speaker uttered in a meeting. Applications of diarization systemsextend to other domains than meetings, for example, lectures, telephone, television, and radio. Besides, diarization enhances the performance of several speech technologies such as speaker recognition, automatic transcription, and speaker tracking. Methodologies previously used in developing diarization systems are discussed. Prior results and techniques are studied and compared. Methods such as Hidden Markov Models and Gaussian Mixture Models that are used in speaker recognition and other speech technologies are also used in speaker diarization. The objective of this thesis is to develop a speaker diarization system in meeting domain. Experimental part of this work indicates that zero-crossing rate can be used effectively in breaking down the audio stream into segments, and adaptive Gaussian Models fit adequately short audio segments. Results show that 35 Gaussian Models and one second as average length of each segment are optimum values to build a diarization system for the tested data. Uniting the segments which are uttered by same speaker is done in a bottom-up clustering by a newapproach of categorizing the mixture weights.

D’abord, ensuite, enfin et 0, De plus: Organisation textuelle par des séries linéaires dans les articles de recherche

The study examines the signalling of text organisation in research articles (RA) in French. The work concentrates on a particular type of organisation provided by text sequences, i.e. structures organising text to items of which at least some are signalled by markers of addition or order: First… 0… The third point… In addition… / Premièrement… 0… Le troisième point… De plus… By indicating the way the text is organised, these structures guide the reader in the reading process so that he doesn’t need to interpret the text structure himself. The aim of the work is to study factors affecting the marking of text sequences. Why is their structure sometimes signalled explicitly by markers such as secondly, whereas in other places such markers are not used? The corpus is manually XML-annotated and consists of 90 RAs (~800 000 words) in French from the fields of linguistics, education and history. The analysis highlights several factors affecting the marking of text sequences. First, exact markers (such as fist ) seem to be more frequent in sequences where all the items are explicitly signalled by a marker, whereas additive markers (such as moreover) are used in sequences with both explicitly signalled and unmarked items. The marking of explicitly signalled sequences seems thus to be precise and even repetitive, whereas the signalling of sequences with unmarked items is altogether more vague. Second, the marking of text sequences seems to depend on the length of the text. The longer the text segment, the more vague the marking. Additive markers and unmarked items are more frequent in longer sequences possibly covering several pages, whereas shorter sequences are often signalled explicitly by exact markers. Also the marker types vary according to the sequence length. Anaphoric expressions, such as first, are fairly close to their referents and are used in short sequences, connectors, such as secondly, are frequently used in sequences of intermediate length, whereas the longest sequences are often signalled by constructions composed of an ordinal and a noun acting as a subject of the sentence: The first item is… Finally, the marking of text organisation depends also on the discipline the RA belongs to. In linguistics, the marking is fairly frequent and precise; exact markers such as second are the most used, and structures with unmarked items are less common. Similarly, the marking is fairly frequent in education. In this field, however, it is also less precise than in linguistics, with frequent unmarked items and additive markers. History, on the other hand, is characterised by less frequent marking. In addition, when used, the marking in this field is also less precise and less explicit.

Anaerobic reactors with biofilter and different diameter-length ratios in cassava starch industry wastewater treatment

The use of anaerobic reactors with media support in the treatment of wastewater from the cassava starch industry has emerged as a viable option because it allows the application of high organic loads and a significant reduction of the HDT needed for the treatment. This research aimed at studying the process of biodigestion in two anaerobic reactors with bamboo support, in the treatment of effluent of cassava starch, by evaluating their performance. The two reactors used present the following diameter: length ratio, 1:6 and 1:3. The organic loads applied to the systems were 0.519, 1.156, 1.471, 3.049, 4.347, 4.708 and 5.601g.L-1.d-1. Regarding the efficiency of removal of COD, TS and TVS, no statistically significant differences were obtained between the reactors. The two systems evaluated showed a stable behavior with respect to the VA/TA (volatile acidity/total alkalinity) for all submitted loads. The reactors tended to the maintenance of biogas production as a function of consumed COD for the last three organic loads applied, indicating an ability to withstand higher organic loads.

Dispersal and metacommunity dynamics in a soft-sediment benthic system : how well is the seafloor connected?

Connectivity depends on rates of dispersal between communities. For marine soft-sediment communities continued small-scale dispersal as post-larvae and as adults can be equally important in maintaining community composition, as initial recruitment of substrate by pelagic larvae. In this thesis post-larval dispersal strategies of benthic invertebrates, as well as mechanisms by which communities are connected were investigated. Such knowledge on dispersal is scarce, due to the difficulties in actually measuring dispersal directly in nature, and dispersal has not previously been quantified in the Baltic Sea. Different trap-types were used underwater to capture dispersing invertebrates at different sites, while in parallel measuring waves and currents. Local community composition was found to change predictably under varying rates of dispersal and physical connectivity (waves and currents). This response was, however, dependent on dispersal-related traits of taxa. Actively dispersing taxa will be relatively better at maintaining their position, as they are not as dependent on hydrodynamic conditions for dispersal and will be less prone to be passively transported by currents. Taxa also dispersed in relative proportions that were distinctly different from resident community composition and a significant proportion (40 %) of taxa were found to lack a planktonic larval life-stage. Community assembly was re-started in a large-scale manipulative field experiment over one year across several sites, which revealed how patterns of community composition (α-, β- and λ-diversity) change depending on rates of dispersal. Results also demonstrated that in response to small-scale disturbance, initial recruitment was by nearby-dominant species after which other species arrived from successively further away. At later assembly time, the number of coexisting species increased beyond what was expected purely by local niche requirements (species sorting), transferring regional differences in community composition (β-diversity) to the local scale (α-diversity, mass effect). Findings of this thesis complement more theoretical studies in metacommunity ecology by demonstrating that understanding how and when individuals disperse relative to underlying environmental heterogeneity is key to interpreting how patterns of diversity change across different spatial scales. Such information from nature is critical when predicting responses to, for example, different types of disturbances or management actions in conservation.

Estimation of Fetal Weight during Labor: Still a Challenge

Objective To evaluate the accuracy of fetal weight prediction by ultrasonography labor employing a formula including the linear measurements of femur length (FL) and mid-thigh soft-tissue thickness (STT). Methods We conducted a prospective study involving singleton uncomplicated term pregnancies within 48 hours of delivery. Only pregnancies with a cephalic fetus admitted in the labor ward for elective cesarean section, induction of labor or spontaneous labor were included. We excluded all non-Caucasian women, the ones previously diagnosed with gestational diabetes and the ones with evidence of ruptured membranes. Fetal weight estimates were calculated using a previously proposed formula [estimated fetal weight = [1] 1687.47 + (54.1 x FL) + (76.68 x STT). The relationship between actual birth weight and estimated fetal weight was analyzed using Pearson's correlation. The formula's performance was assessed by calculating the signed and absolute errors. Mean weight difference and signed percentage error were calculated for birth weight divided into three subgroups: < 3000 g; 3000-4000g; and > 4000 g. Results We included for analysis 145 cases and found a significant, yet low, linear relationship between birth weight and estimated fetal weight (p < 0.001; R2 = 0.197) with an absolute mean error of 10.6%. The lowest mean percentage error (0.3%) corresponded to the subgroup with birth weight between 3000 g and 4000 g. Conclusions This study demonstrates a poor correlation between actual birth weight and the estimated fetal weight using a formula based on femur length and mid-thigh soft-tissue thickness, both linear parameters. Although avoidance of circumferential ultrasound measurements might prove to be beneficial, it is still yet to be found a fetal estimation formula that can be both accurate and simple to perform.

"It's Barack Obama Who's Soft" Political Bias on BBC, CNN, and Fox News: A Critical Discourse Analysis of News Websites

Tutkielmassa analysoidaan kolmen internetsivuston uutisartikkeleita kielitieteen näkökulmasta. Tavoitteena on selvittää esiintyykö internetsivustojen BBC, CNN ja Fox News uutisoinnissa politiikkaan liittyviä ennakkoasenteita tai puolueellisuuksia ja miten ne käytännössä näkyvät uutisartikkeleiden kielessä. Kriittiseen diskurssianalyysiin pohjautuen tutkielma esittelee jokaisen uutissivuston taustaan (esimerkiksi rakenteeseen ja rahoitukseen) liittyviä seikkoja sekä mediadiskurssiin ja politiikkaan liittyvät taustatiedot, jolla taataan Norman Fairclough'n kolmivaiheisen menetelmän mahdollisimman perusteellinen toteuttaminen. Uutissivustoja analysoidaan kriittiselle diskurssianalyysille sopivan funktionaalisen kieliopin ja muiden lingvististen välineiden avulla. Koko aineiston (404 artikkelia) otsikot analysoidaan ensin, minkä jälkeen analysoidaan yhdeksän kokonaista artikkelia kolmeen eri aihealueeseen liittyen niin, että jokaiselta internetsivustolta analysoidaan yksi artikkeli jokaista aihetta kohden. Analyysikeinoina käytetään ensisijaisesti systeemis-funktionaalisen kieliopin tekstuaalisen metafunktion välineitä (thematic structure). Myös ideationaalisen metafunktion välineitä (transitivity), referenssiketjuja (referential identity chains) ja leksikaalista analyysia käytetään hyväksi. Lähtökohtaisesti tavoitteena on analysoida uutissivustoja vertailevasti, jolloin analyysin tulokset ovat paremmin havainnoitavissa ja perusteltavissa. Hypoteesi aikaisempien tutkimusten ja yleisen mielikuvan perusteella on, että CNN uutisoi demokraattipuolueelle ja Fox News taas republikaanipuolueelle edulliseen sävyyn. Tutkimustulokset vaihtelivat hypoteesia tukevista ja sen vastaisista tuloksista niihin, jotka eivät olleet tarpeeksi tuettuja kumpaankaan suuntaan. Vahvimmat tulokset ovat kuitenkin hypoteesia tukevia, joten tässä tutkielmassa todetaan, ettei uutisointi ole puolueetonta ainakaan näiden kolmen internetsivuston kohdalla. Lisäksi muutaman aihealueen kohdalla uutisointi on niin toistuvaa tietystä näkökulmasta, että luonnollistumisteorian mukaista aatteiden luonnollistumista saattaa tapahtua. Tutkielmassa käytettyjen menetelmien menestyksen perusteella suositellaan, että tekstuaalisen metafunktion analyysivälineitä käytetään enemmän. Lisäksi suositellaan meta-analyysin harkitsemista, jotta voitaisiin selvittää, mitkä analyysimetodit parhaiten sopivat minkäkinlaisen aineiston analysointiin.

A and B mode ultrasonography in preoperative evaluation of lens and posterior segment of dogs eyes with cataract

Ultrasonography of the lens and posterior segment is an indispensable step in the preoperative evaluation of dogs with cataracts, since ophthalmoscopy is not feasible when there is opacification of the lens. This study evaluated the echographic conditions of cataractous lens and fundus of the eye in dogs affected by cataracts. The study was conducted in 30 dogs (56 eyes), 10 males and 20 females, with different types of cataracts at different stages of development. Echography in A and B modes, simultaneously, was carried out for the examination of the lens and posterior segment. The examinations revealed anterior cortical, posterior cortical and nuclear cataract in 12 eyes (21.4%), anterior cortical, posterior cortical, nuclear and posterior capsular in 23 eyes (41%), anterior cortical, posterior cortical and posterior capsular cataract in one eye (1.7%), anterior cortical and nuclear cataract in one eye (1.7%), anterior cortical, nuclear and posterior capsular cataract in five eyes (8.9%), and anterior cortical cataract in seven eyes (12.5%). Abnormal ultrasonographic alterations were observed in the posterior segment in 26 eyes evaluated (46.4%). Vitreal degeneration was detected in 12 eyes (21.4%), images of vitreal exudate or hemorrhage in seven eyes (12.5%), persistence of hyaloid artery in four eyes (7.1%) and lens subluxation in three eyes (5.3%). The results obtained reiterate the importance of ultrasonography in canine patients presented for cataract surgery given that alterations of the posterior segment are difficult to identify in a clinical examination when the lens is opacified.

A Soft Contact Collision Method For Real-Time Simulation of Triangularized Geometries in Multibody Dynamics

When modeling machines in their natural working environment collisions become a very important feature in terms of simulation accuracy. By expanding the simulation to include the operation environment, the need for a general collision model that is able to handle a wide variety of cases has become central in the development of simulation environments. With the addition of the operating environment the challenges for the collision modeling method also change. More simultaneous contacts with more objects occur in more complicated situations. This means that the real-time requirement becomes more difficult to meet. Common problems in current collision modeling methods include for example dependency on the geometry shape or mesh density, calculation need increasing exponentially in respect to the number of contacts, the lack of a proper friction model and failures due to certain configurations like closed kinematic loops. All these problems mean that the current modeling methods will fail in certain situations. A method that would not fail in any situation is not very realistic but improvements can be made over the current methods.

Synchrony is more than overlap: measuring phenological synchronization considering time length and intensity

The degree of flowering and fruiting synchronization is believed to have ecological and evolutionary relevance at several scales. Here we discuss some measures that have been used to estimate synchrony and propose an index that incorporates both the entire length of an individual phenophase and variation in the number of flowers or fruits over that time period. This new index describes more accurately the phenological synchrony among individuals and populations.

Construction of occluded recombinant baculoviruses containing the full-length cry1Ab and cry1Ac genes from Bacillus thuringiensis

The administration of baculoviruses to insects for bioassay purposes is carried out, in most cases, by contamination of food surfaces with a known amount of occlusion bodies (OBs). Since per os infection is the natural route of infection, occluded recombinant viruses containing crystal protein genes (cry1Ab and cry1Ac) from Bacillus thuringiensis were constructed for comparison with the baculovirus prototype Autographa californica nucleopolyhedrovirus (AcNPV). The transfer vector pAcUW2B was used for construction of occluded recombinant viruses. The transfer vector containing the crystal protein genes was cotransfected with linearized DNA from a non-occluded recombinant virus. The isolation of recombinant viruses was greatly facilitated by the reduction of background "wild type" virus and the increased proportion of recombinant viruses. Since the recombinant viruses containing full-length and truncated forms of the crystal protein genes did not seem to improve the pathogenicity of the recombinant viruses when compared with the wild type AcNPV, and in order to compare expression levels of the full-length crystal proteins produced by non-occluded and occluded recombinant viruses the full-length cry1Ab and cry1Ac genes were chosen for construction of occluded recombinant viruses. The recombinant viruses containing full-length and truncated forms of the crystal protein genes did not seem to improve its pathogenicity but the size of the larvae infected with the recombinant viruses was significantly smaller than that of larvae infected with the wild type virus.

Characterization of six rat strains (Rattus norvegicus) by mitochondrial DNA restriction fragment length polymorphism

Restriction fragment length polymorphism (RFLP) was used to examine the extent of mtDNA polymorphism among six strains of rats (Rattus norvegicus) - Wistar, Wistar Munich, Brown Norway, Wistar Kyoto, SHR and SHR-SP. A survey of 26 restriction enzymes has revealed a low level of genetic divergence among strains. The sites of cleavage by EcoRI, NcoI and XmnI were shown to be polymorphic. The use of these three enzymes allows the 6 strains to be classified into 4 haplotypes and identifies specific markers for each one. The percentage of sequence divergence among all pairs of haplotypes ranged from 0.035 to 0.33%, which is the result of a severe population constriction undergone by the strains. These haplotypes are easily demonstrable and therefore RFLP analysis can be employed for genetic monitoring of rats within animal facilities or among different laboratories.

Effect of practice and span length on the dual-task coordination executive test

The measure "mu", proposed as an index of the ability to coordinate concurrent box-crossing (BC) and digit-span (DS) tasks in the dual task (DT), should reflect the capacity of the executive component of the working memory system. We investigated the effect of practice in BC and of a change in the digit span on mu by adding previous practice trials in BC and diminishing, maintaining or increasing the digit sequence length. The mu behavior was evaluated throughout three trials of the test. Reported strategies in digit tasks were also analyzed. Subjects with diminished span showed the best performance in DT due to a stable performance in DS and BC in the single- and dual-task conditions. These subjects also showed a more stable performance throughout trials. Subjects with diminished span tended to employ effortless strategies, whereas subjects with increased span employed effort-requiring strategies and showed the lowest means of mu. Subjects with initial practice trials showed the best performance in BC and the most differentiated performance between the single- and dual-task conditions in BC. The correlation coefficient between the mu values obtained in the first and second trials was 0.814 for subjects with diminished span and practice trials in BC. It seems that the within-session practice in BC and the performance variability in DS affect the reliability of the index mu. To control these factors we propose the introduction of previous practice trials in BC and a modification of the current method to determine the digit sequence length. This proposal should contribute to the development of a more reliable method to evaluate the executive capacity of coordination in the dual-task paradigm.

Management of Soft Information in Sales and Operations Planning

With a Sales and Operations Planning (S&OP) process, a company aims to manage the demand and supply by planning and forecasting. The studied company uses an integrated S&OP process to improve the company's operations. The aim of this thesis is to develop this business process by finding the best possible way to manage the soft information in S&OP, whilst also understanding the importance and types (assumptions, risks and opportunities) of soft information in S&OP. The soft information in S&OP helps to refine future S&OP planning, taking into account the uncertainties that affect the balance of the long-term demand and supply (typically 12-18 months). The literature review was used to create a framework for soft information management process in S&OP. There were not found a concrete way how to manage soft information in the existing literature. In consequence of the poor literature available the Knowledge Management literature was used as the base for the framework creation, which was seen in the very same type of information management like the soft information management is. The framework created a four-stage process to manage soft information in S&OP that included also the required support systems. First phase is collecting and acquiring soft information in S&OP, which include also categorization. The categorization was the cornerstone to identify different requirements that needs to be taken into consideration when managing soft information in S&OP process. The next phase focus on storing data, which purpose is to ensure the soft information is managed in a common system (support system) in a way that the following phase makes it available to users in S&OP who need by help of sharing and applications process. The last phase target is to use the soft information to understand assumptions and thoughts of users behind the numbers in S&OP plans. With this soft management process the support system will have a key role. The support system, like S&OP tool, ensures that soft information is stored in the right places, kept up-to-date and relevancy. The soft information management process in S&OP strives to improve the relevant soft information documenting behind the S&OP plans into the S&OP support system. The process offers an opportunity to individuals to review, comment and evaluate soft information in S&OP made by their own or others. In the case company it was noticed that without a properly documented and distributed soft information in S&OP it was seen to cause mistrust towards the planning.

Association between EcoRI fragment-length polymorphism of the immunoglobulin lambda variable 8 (IGLV8) gene family with rheumatoid arthritis and systemic lupus erythematosus

The human immunoglobulin lambda variable 8 (IGLV8) subgroup is a gene family containing three members, one of them included in a monomorphic 3.7-kb EcoRI genomic fragment located at the major lambda variable locus on chromosome 22q11.1 (gene IGLV8a, EMBL accession No. Z73650) at 100% frequency in the normal urban population. The second is a polymorphic RFLP allele included in a 6.0-kb EcoRI fragment at 10% frequency, and the third is located in a monomorphic 8.0-kb EcoRI fragment at 100% frequency, the last being translocated to chromosome 8q11.2 and considered to be an orphan gene. Our Southern blot-EcoRI-RFLP studies in normal individuals and in patients with rheumatoid arthritis (RA) or with systemic lupus erythematosus (SLE), using a specific probe for the IGLV8 gene family (probe pVL8, EMBL accession No. X75424), have revealed the two monomorphic genomic fragments containing the IGLV8 genes, i.e., the 3.7-kb fragment from chromosome 22q11.1 and the 8.0-kb fragment from 8q11.2, both occurring at 100% frequency (103 normal individuals, 48 RA and 28 SLE patients analyzed), but absence of the 6.0-kb IGLV8 polymorphic RFLP allele in all RA or SLE patients. As expected, the frequency of the 6.0-kb allele among the normal individuals was 10%. These findings suggest an association between the absence of the 6.0-kb EcoRI fragment and rheumatoid arthritis and systemic lupus erythematosus.