Effective 20 newsgroups dataset cleaning

The rapid increase in the number of text documents available on the Internet has created pressure to use effective cleaning techniques. Cleaning techniques are needed for converting these documents to structured documents. Text cleaning techniques are one of the key mechanisms in typical text mining application frameworks. In this paper, we explore the role of text cleaning in the 20 newsgroups dataset, and report on experimental results.

Overview of the 2015 ALTA shared task: Identifying French cognates in English text

This paper presents an overview of the 6th ALTA shared task that ran in 2015. The task was to identify in English texts all the potential cognates from the perspective of the French language. In other words, identify all the words in the English text that would acceptably translate into a similar word in French. We present the motivations for the task, the description of the data and the results of the 4 participating teams. We discuss the results against a baseline and prior work.

An effective approach to verbose queries using a limited dependencies language model

Intuitively, any `bag of words' approach in IR should benefit from taking term dependencies into account. Unfortunately, for years the results of exploiting such dependencies have been mixed or inconclusive. To improve the situation, this paper shows how the natural language properties of the target documents can be used to transform and enrich the term dependencies to more useful statistics. This is done in three steps. The term co-occurrence statistics of queries and documents are each represented by a Markov chain. The paper proves that such a chain is ergodic, and therefore its asymptotic behavior is unique, stationary, and independent of the initial state. Next, the stationary distribution is taken to model queries and documents, rather than their initial distri- butions. Finally, ranking is achieved following the customary language modeling paradigm. The main contribution of this paper is to argue why the asymptotic behavior of the document model is a better representation then just the document's initial distribution. A secondary contribution is to investigate the practical application of this representation in case the queries become increasingly verbose. In the experiments (based on Lemur's search engine substrate) the default query model was replaced by the stable distribution of the query. Just modeling the query this way already resulted in significant improvements over a standard language model baseline. The results were on a par or better than more sophisticated algorithms that use fine-tuned parameters or extensive training. Moreover, the more verbose the query, the more effective the approach seems to become.

Exponentiated gradient algorithms for conditional random fields and max-margin Markov Networks

Log-linear and maximum-margin models are two commonly-used methods in supervised machine learning, and are frequently used in structured prediction problems. Efficient learning of parameters in these models is therefore an important problem, and becomes a key factor when learning from very large data sets. This paper describes exponentiated gradient (EG) algorithms for training such models, where EG updates are applied to the convex dual of either the log-linear or max-margin objective function; the dual in both the log-linear and max-margin cases corresponds to minimizing a convex function with simplex constraints. We study both batch and online variants of the algorithm, and provide rates of convergence for both cases. In the max-margin case, O(1/ε) EG updates are required to reach a given accuracy ε in the dual; in contrast, for log-linear models only O(log(1/ε)) updates are required. For both the max-margin and log-linear cases, our bounds suggest that the online EG algorithm requires a factor of n less computation to reach a desired accuracy than the batch EG algorithm, where n is the number of training examples. Our experiments confirm that the online algorithms are much faster than the batch algorithms in practice. We describe how the EG updates factor in a convenient way for structured prediction problems, allowing the algorithms to be efficiently applied to problems such as sequence learning or natural language parsing. We perform extensive evaluation of the algorithms, comparing them to L-BFGS and stochastic gradient descent for log-linear models, and to SVM-Struct for max-margin models. The algorithms are applied to a multi-class problem as well as to a more complex large-scale parsing task. In all these settings, the EG algorithms presented here outperform the other methods.

An effective approach to verbose queries using a limited dependencies language model

Intuitively, any ‘bag of words’ approach in IR should benefit from taking term dependencies into account. Unfortunately, for years the results of exploiting such dependencies have been mixed or inconclusive. To improve the situation, this paper shows how the natural language properties of the target documents can be used to transform and enrich the term dependencies to more useful statistics. This is done in three steps. The term co-occurrence statistics of queries and documents are each represented by a Markov chain. The paper proves that such a chain is ergodic, and therefore its asymptotic behavior is unique, stationary, and independent of the initial state. Next, the stationary distribution is taken to model queries and documents, rather than their initial distributions. Finally, ranking is achieved following the customary language modeling paradigm. The main contribution of this paper is to argue why the asymptotic behavior of the document model is a better representation then just the document’s initial distribution. A secondary contribution is to investigate the practical application of this representation in case the queries become increasingly verbose. In the experiments (based on Lemur’s search engine substrate) the default query model was replaced by the stable distribution of the query. Just modeling the query this way already resulted in significant improvements over a standard language model baseline. The results were on a par or better than more sophisticated algorithms that use fine-tuned parameters or extensive training. Moreover, the more verbose the query, the more effective the approach seems to become.

vSpeak: Edge detection based feature extraction for sign to text conversion

This paper presents 'vSpeak', the first initiative taken in Pakistan for ICT enabled conversion of dynamic Sign Urdu gestures into natural language sentences. To realize this, vSpeak has adopted a novel approach for feature extraction using edge detection and image compression which gives input to the Artificial Neural Network that recognizes the gesture. This technique caters for the blurred images as well. The training and testing is currently being performed on a dataset of 200 patterns of 20 words from Sign Urdu with target accuracy of 90% and above.

Neurolinguistic contributions to understanding the bilingual mental lexicon

Introduction Many bilinguals will have had the experience of unintentionally reading something in a language other than the intended one (e.g. MUG to mean mosquito in Dutch rather than a receptacle for a hot drink, as one of the possible intended English meanings), of finding themselves blocked on a word for which many alternatives suggest themselves (but, somewhat annoyingly, not in the right language), of their accent changing when stressed or tired and, occasionally, of starting to speak in a language that is not understood by those around them. These instances where lexical access appears compromised and control over language behavior is reduced hint at the intricate structure of the bilingual lexical architecture and the complexity of the processes by which knowledge is accessed and retrieved. While bilinguals might tend to blame word finding and other language problems on their bilinguality, these difficulties per se are not unique to the bilingual population. However, what is unique, and yet far more common than is appreciated by monolinguals, is the cognitive architecture that subserves bilingual language processing. With bilingualism (and multilingualism) the rule rather than the exception (Grosjean, 1982), this architecture may well be the default structure of the language processing system. As such, it is critical that we understand more fully not only how the processing of more than one language is subserved by the brain, but also how this understanding furthers our knowledge of the cognitive architecture that encapsulates the bilingual mental lexicon. The neurolinguistic approach to bilingualism focuses on determining the manner in which the two (or more) languages are stored in the brain and how they are differentially (or similarly) processed. The underlying assumption is that the acquisition of more than one language requires at the very least a change to or expansion of the existing lexicon, if not the formation of language-specific components, and this is likely to manifest in some way at the physiological level. There are many sources of information, ranging from data on bilingual aphasic patients (Paradis, 1977, 1985, 1997) to lateralization (Vaid, 1983; see Hull & Vaid, 2006, for a review), recordings of event-related potentials (ERPs) (e.g. Ardal et al., 1990; Phillips et al., 2006), and positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) studies of neurologically intact bilinguals (see Indefrey, 2006; Vaid & Hull, 2002, for reviews). Following the consideration of methodological issues and interpretative limitations that characterize these approaches, the chapter focuses on how the application of these approaches has furthered our understanding of (1) selectivity of bilingual lexical access, (2) distinctions between word types in the bilingual lexicon and (3) control processes that enable language selection.

Losing their marbles : syntax-free programming for assessing problem-solving skills

Novice programmers have difficulty developing an algorithmic solution while simultaneously obeying the syntactic constraints of the target programming language. To see how students fare in algorithmic problem solving when not burdened by syntax, we conducted an experiment in which a large class of beginning programmers were required to write a solution to a computational problem in structured English, as if instructing a child, without reference to program code at all. The students produced an unexpectedly wide range of correct, and attempted, solutions, some of which had not occurred to their teachers. We also found that many common programming errors were evident in the natural language algorithms, including failure to ensure loop termination, hardwiring of solutions, failure to properly initialise the computation, and use of unnecessary temporary variables, suggesting that these mistakes are caused by inexperience at thinking algorithmically, rather than difficulties in expressing solutions as program code.

Detecting and resolving redundancies in EP3P policies

Current regulatory requirements on data privacy make it increasingly important for enterprises to be able to verify and audit their compliance with their privacy policies. Traditionally, a privacy policy is written in a natural language. Such policies inherit the potential ambiguity, inconsistency and mis-interpretation of natural text. Hence, formal languages are emerging to allow a precise specification of enforceable privacy policies that can be verified. The EP3P language is one such formal language. An EP3P privacy policy of an enterprise consists of many rules. Given the semantics of the language, there may exist some rules in the ruleset which can never be used, these rules are referred to as redundant rules. Redundancies adversely affect privacy policies in several ways. Firstly, redundant rules reduce the efficiency of operations on privacy policies. Secondly, they may misdirect the policy auditor when determining the outcome of a policy. Therefore, in order to address these deficiencies it is important to identify and resolve redundancies. This thesis introduces the concept of minimal privacy policy - a policy that is free of redundancy. The essential component for maintaining the minimality of privacy policies is to determine the effects of the rules on each other. Hence, redundancy detection and resolution frameworks are proposed. Pair-wise redundancy detection is the central concept in these frameworks and it suggests a pair-wise comparison of the rules in order to detect redundancies. In addition, the thesis introduces a policy management tool that assists policy auditors in performing several operations on an EP3P privacy policy while maintaining its minimality. Formal results comparing alternative notions of redundancy, and how this would affect the tool, are also presented.

A study and comparison of multimedia Web searching : 1997-2006

Searching for multimedia is an important activity for users of Web search engines. Studying user's interactions with Web search engine multimedia buttons, including image, audio, and video, is important for the development of multimedia Web search systems. This article provides results from a Weblog analysis study of multimedia Web searching by Dogpile users in 2006. The study analyzes the (a) duration, size, and structure of Web search queries and sessions; (b) user demographics; (c) most popular multimedia Web searching terms; and (d) use of advanced Web search techniques including Boolean and natural language. The current study findings are compared with results from previous multimedia Web searching studies. The key findings are: (a) Since 1997, image search consistently is the dominant media type searched followed by audio and video; (b) multimedia search duration is still short (>50% of searching episodes are <1 min), using few search terms; (c) many multimedia searches are for information about people, especially in audio search; and (d) multimedia search has begun to shift from entertainment to other categories such as medical, sports, and technology (based on the most repeated terms). Implications for design of Web multimedia search engines are discussed.

Clustered blind beamforming from ad-hoc microphone arrays

Microphone arrays have been used in various applications to capture conversations, such as in meetings and teleconferences. In many cases, the microphone and likely source locations are known \emph{a priori}, and calculating beamforming filters is therefore straightforward. In ad-hoc situations, however, when the microphones have not been systematically positioned, this information is not available and beamforming must be achieved blindly. In achieving this, a commonly neglected issue is whether it is optimal to use all of the available microphones, or only an advantageous subset of these. This paper commences by reviewing different approaches to blind beamforming, characterising them by the way they estimate the signal propagation vector and the spatial coherence of noise in the absence of prior knowledge of microphone and speaker locations. Following this, a novel clustered approach to blind beamforming is motivated and developed. Without using any prior geometrical information, microphones are first grouped into localised clusters, which are then ranked according to their relative distance from a speaker. Beamforming is then performed using either the closest microphone cluster, or a weighted combination of clusters. The clustered algorithms are compared to the full set of microphones in experiments on a database recorded on different ad-hoc array geometries. These experiments evaluate the methods in terms of signal enhancement as well as performance on a large vocabulary speech recognition task.

The delta-phase spectrum with application to voice activity detection and speaker recognition

For several reasons, the Fourier phase domain is less favored than the magnitude domain in signal processing and modeling of speech. To correctly analyze the phase, several factors must be considered and compensated, including the effect of the step size, windowing function and other processing parameters. Building on a review of these factors, this paper investigates a spectral representation based on the Instantaneous Frequency Deviation, but in which the step size between processing frames is used in calculating phase changes, rather than the traditional single sample interval. Reflecting these longer intervals, the term delta-phase spectrum is used to distinguish this from instantaneous derivatives. Experiments show that mel-frequency cepstral coefficients features derived from the delta-phase spectrum (termed Mel-Frequency delta-phase features) can produce broadly similar performance to equivalent magnitude domain features for both voice activity detection and speaker recognition tasks. Further, it is shown that the fusion of the magnitude and phase representations yields performance benefits over either in isolation.

Discriminative optimisation of the figure of merit for phonetic spoken term detection

This work proposes to improve spoken term detection (STD) accuracy by optimising the Figure of Merit (FOM). In this article, the index takes the form of phonetic posterior-feature matrix. Accuracy is improved by formulating STD as a discriminative training problem and directly optimising the FOM, through its use as an objective function to train a transformation of the index. The outcome of indexing is then a matrix of enhanced posterior-features that are directly tailored for the STD task. The technique is shown to improve the FOM by up to 13% on held-out data. Additional analysis explores the effect of the technique on phone recognition accuracy, examines the actual values of the learned transform, and demonstrates that using an extended training data set results in further improvement in the FOM.