The art and science of audio book production.

Shipping list no.: 96-0048-P.

A talking computer terminal /

Thesis--University of Illinois at Urbana-Champaign.

A novel framework of training hidden Markov support vector machines from lightly-annotated data

Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18% in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3% and 10.6% being achieved.

Learning higher-level features with convolutional restricted Boltzmann machines for sentiment analysis

In recent years, learning word vector representations has attracted much interest in Natural Language Processing. Word representations or embeddings learned using unsupervised methods help addressing the problem of traditional bag-of-word approaches which fail to capture contextual semantics. In this paper we go beyond the vector representations at the word level and propose a novel framework that learns higher-level feature representations of n-grams, phrases and sentences using a deep neural network built from stacked Convolutional Restricted Boltzmann Machines (CRBMs). These representations have been shown to map syntactically and semantically related n-grams to closeby locations in the hidden feature space. We have experimented to additionally incorporate these higher-level features into supervised classifier training for two sentiment analysis tasks: subjectivity classification and sentiment classification. Our results have demonstrated the success of our proposed framework with 4% improvement in accuracy observed for subjectivity classification and improved the results achieved for sentiment classification over models trained without our higher level features.

It's about bits of machines & stuff:why girls don't 'do' engineering

This paper draws upon part of the findings of an ethnographic study in which two seventeen year old girls were employed to interview their peer about engineering as a study and career choice. It argues that whilst girls do view engineering as being generally masculine in nature, other factors such as a lack of female role models and an emphasis on physics and maths act as barriers to young women entering the discipline. The paper concludes by noting that engineering has much to offer young women, the problem is, they simply don't know this is the case! Copyright © 2013 Jane Andrews & Robin Clark.

It’s about bits of machines & stuff:why girls don’t ‘do’ engineering

This paper draws upon part of the findings of an ethnographic study in which two seventeen year old girls were employed to interview their peer about engineering as a study and career choice. It argues that whilst girls do view engineering as being generally masculine in nature, other factors such as a lack of female role models and an emphasis on physics and maths act as barriers to young women entering the discipline. The paper concludes by noting that engineering has much to offer young women, the problem is, they simply don’t know this is the case!

Unsupervised parameter estimation for one-class support vector machines

Although the hyper-plane based One-Class Support Vector Machine (OCSVM) and the hyper-spherical based Support Vector Data Description (SVDD) algorithms have been shown to be very effective in detecting outliers, their performance on noisy and unlabeled training data has not been widely studied. Moreover, only a few heuristic approaches have been proposed to set the different parameters of these methods in an unsupervised manner. In this paper, we propose two unsupervised methods for estimating the optimal parameter settings to train OCSVM and SVDD models, based on analysing the structure of the data. We show that our heuristic is substantially faster than existing parameter estimation approaches while its accuracy is comparable with supervised parameter learning methods, such as grid-search with crossvalidation on labeled data. In addition, our proposed approaches can be used to prepare a labeled data set for a OCSVM or a SVDD from unlabeled data.

Non-invasive winding fault detection for induction machines based on stray flux magnetic sensors

Non-intrusive monitoring of health state of induction machines within industrial process and harsh environments poses a technical challenge. In the field, winding failures are a major fault accounting for over 45% of total machine failures. In the literature, many condition monitoring techniques based on different failure mechanisms and fault indicators have been developed where the machine current signature analysis (MCSA) is a very popular and effective method at this stage. However, it is extremely difficult to distinguish different types of failures and hard to obtain local information if a non-intrusive method is adopted. Typically, some sensors need to be installed inside the machines for collecting key information, which leads to disruption to the machine operation and additional costs. This paper presents a new non-invasive monitoring method based on GMRs to measure stray flux leaked from the machines. It is focused on the influence of potential winding failures on the stray magnetic flux in induction machines. Finite element analysis and experimental tests on a 1.5-kW machine are presented to validate the proposed method. With time-frequency spectrogram analysis, it is proven to be effective to detect several winding faults by referencing stray flux information. The novelty lies in the implement of GMR sensing and analysis of machine faults.