223 research outputs found
High-Level Approaches to Confidence Estimation in Speech Recognition
No full textWe describe some high-level approaches to estimating confidence scores for the words output by a speech recognizer. By "high-level" we mean that the proposed measures do not rely on decoder specific "side information" and so should find more general applicability than measures that have been developed for specific recognizers. Our main approach is to attempt to decouple the language modeling and acoustic modeling in the recognizer in order to generate independent information from these two sources that can then be used for estimation of confidence. We isolate these two information sources by using a phone recognizer working in parallel with the word recognizer. A set of techniques for estimating confidence measures using the phone recognizer output in conjunction with the word recognizer output is described. The most effective of these techniques is based on the construction of "metamodels," which generate alternative word hypotheses for an utterance. An alternative approach requires no other recognizers or extra information for confidence estimation and is based on the notion that a word that is semantically "distant" from the other decoded words in the utterance is likely to be incorrect. We describe a method for constructing "semantic similarities" between words and hence estimating a confidence. Results using the U.K. version of the Wall Street Journal are given for each technique
Recommender Systems for the Semantic Web
No full textThis paper presents a semantic approach to Recommender Systems (RS), to exploit available contextual information about both the items to be recommended and the recommendation process, in an attempt to overcome some of the shortcomings of traditional RS implementations. An ontology is used as a backbone to the system in the proposed architecture to represent the problem domain, while multiple web services are orchestrated to compose a suitable recommendation model, matching the current recommendation context at run-time. In order to allow for such dynamic behaviour, the proposed system tackles the recommendation problem by applying existing RS techniques on three different levels: the selection of appropriate sets of features, recommendation model and recommendable items
Secure Interaction Models for the HealthAgents System
No full textDistributed decision support systems designed for healthcare use can benefit from services and information available across a decentralised environment. The sophisticated nature of collaboration among involved partners who contribute services or sensitive data in this paradigm, however, demands careful attention from the beginning of designing such systems. Apart from the traditional need of secure data transmission across clinical centres, a more important issue arises from the need of consensus for access to system-wide resources by separately managed user groups from each centre. A primary concern is the determination of interactive tasks that should be made available to authorised users, and further the clinical resources that can be populated into interactions in compliance with user clinical roles and policies. To this end, explicit interaction modelling is put forward along with the contextual constraints within interactions that together enforce secure access, the interaction participation being governed by system-wide policies and local resource access being governed by node-wide policies. Clinical security requirements are comprehensively analysed, prior to the design and building of our security model. The application of the approach results in a Multi-Agent System driven by secure interaction models. This is illustrated using a prototype of the HealthAgents system
Meta-models for Confidence Estimation in Speech Recognition
No full textWe describe an approach to confidence estimation that attempts to decouple the contributions of the acoustic and language model components to speech recognition output. The output of the acoustic models when decoding phonemes is itself modelled using HMMs to produce a set of models which we term meta-models. When benchmarked against a “standard” method for assigning confidence (the N-best score), the meta-models gave a relative improvement of 6.2%. Furthermore, it appears that the N-best and meta-models techniques are complementary, because they tend to fail on different word
A Conceptual Graph Description of Medical data for Brain Tumour Classification
No full textHealthAgents proposes an agent-based distributed decision support system for brain tumour diagnosis and prognosis which employs Magnetic Resonance Imaging and Magnetic Resonance Spectroscopy techniques and genomic profiles. From a knowledge representation view point the distributed nature and the heterogeneity of the data to be integrated pose a number of challenging problems. This paper shows how Conceptual Graphs can be employed to describe the data sources in the HealthAgents system. Such knowledge representation based description of data allows for reasoning power when querying and for data modularisation capabilities
MIAKT: Combining Grid and Web Services for Collaborative Medical Decision Making
No full textProviding semantic web technologies in a medical domain has its obvious advantages. Having distributed services using shared domain vocabularies provides a great impetus for the integration of disparate hospital information sytems, as well as the possibility of providing more accurate diagnoses and a well organised knowledge base for sharing, tutoring and researching. Using such disparate systems requires careful consideration both technicaly, medically and ethically. This paper describes the way the ssytem we have developed offers evidence of the promise of such enhancements in the specific area of screening for breast cancer
Oscillatory dynamics in a double activator motif
No full textBiological oscillators, specially those that constitute the circadian clock, have been extensively modelled as coupled feedback loops of positive and negative elements in gene regulatory circuits. The presence of a negative feedback loop is a necessary condition for the onset of oscillations, and different mechanistic implementations have been modelled, such as transcriptional repression (such as CRY/PER down-regulating activation by CLOCK:BMAL1 in the mammalian clock) or by enhancing proteolysis. It is also a characteristic feature of transcriptional control in clock systems that conserved cis-regulatory sequences (such as E- and D-boxes) are competitively bound by transctiption factors. Using this feature, we propose a new duplicated autoregulated motif where competition for the same promoters by differentially activating transctiption factors drives oscillations
Correlated fluctuations probe dynamics of transcriptional regulation
No full textThe dynamics of transcriptional control involves small numbers of molecules and result in significant fluctuations in protein and mRNA concentrations. The correlations between these intrinsic fluctuations then offer, via the fluctuation dissipation relation, the possibility of capturing the system’s response to external perturbations, and hence the nature of the regulatory activity itself. We study time-dependent noise correlations in simple networks of activators and repressors, varying the topology of causal influence, and using different mechanisms or parameter choices. The distinct correlated fluctuations could be used as signatures for mechanism identification. To that end, we present analytical and numerical results on peaks and delays in correlations between proteins within networks, and the dependence of these features on parameter and mechanism
The Semantic Logger: Supporting Service Building from Personal Context
No full textThe Semantic Logger SL) is presentedas a system for the importing, housing, and exploiting of personal information. The system has been implemented using a number of Semantic Web enabling technologies, and attempts to store the information in a manner adhering to as many W3C recommendations as possible. The Semantic Logger's utility is grounded in two context-based applications, namely a recommender system, and a photo-annotation tool
Spectrum and completeness of the 3 state superintegrable chiral Potts model
No full textWe find the rules which count the energy levels of the three-state superintegrable chiral Potts model and demonstrate that these rules are complete. In the massive phase we show that the spectrum can be written in terms of two single-particle levels where one of the levels has the unusual property that it exists only for a limited range of momentum. We also discuss the relation of the counting rules to the S = 1 XXZ spin chain with anisotropy gamma = 3
- …
