arXiv subjects

Frank Rudzicz

Publications and source records attributed to Frank Rudzicz.

104 records · Page 6Linked to original sources

arXiv2019

Centroid-based deep metric learning for speaker recognition

Speaker embedding models that utilize neural networks to map utterances to a space where distances reflect similarity between speakers have driven recent progress in the speaker recognition task. However, there is still a significant performance gap between recognizing speakers in the training set and unseen speakers. The latter case corresponds to the few-shot learning task, where a trained model is evaluated on unseen classes. Here, we optimize a speaker embedding model with prototypical network loss (PNL), a state-of-the-art approach for the few-shot image classification task. The resulting embedding model outperforms the state-of-the-art triplet loss based models in both speaker verification and identification tasks, for both seen and unseen speakers.

cs.LG↗

arXiv2018

DeepConsensus: using the consensus of features from multiple layers to attain robust image classification

We consider a classifier whose test set is exposed to various perturbations that are not present in the training set. These test samples still contain enough features to map them to the same class as their unperturbed counterpart. Current architectures exhibit rapid degradation of accuracy when trained on standard datasets but then used to classify perturbed samples of that data. To address this, we present a novel architecture named DeepConsensus that significantly improves generalization to these test-time perturbations. Our key insight is that deep neural networks should directly consider summaries of low and high level features when making classifications. Existing convolutional neural networks can be augmented with DeepConsensus, leading to improved resistance against large and small perturbations on MNIST, EMNIST, FashionMNIST, CIFAR10 and SVHN datasets.

cs.CV↗

arXiv2018

Robustness against the channel effect in pathological voice detection

Many people are suffering from voice disorders, which can adversely affect the quality of their lives. In response, some researchers have proposed algorithms for automatic assessment of these disorders, based on voice signals. However, these signals can be sensitive to the recording devices. Indeed, the channel effect is a pervasive problem in machine learning for healthcare. In this study, we propose a detection system for pathological voice, which is robust against the channel effect. This system is based on a bidirectional LSTM network. To increase the performance robustness against channel mismatch, we integrate domain adversarial training (DAT) to eliminate the differences between the devices. When we train on data recorded on a high-quality microphone and evaluate on smartphone data without labels, our robust detection system increases the PR-AUC from 0.8448 to 0.9455 (and 0.9522 with target sample labels). To the best of our knowledge, this is the first study applying unsupervised domain adaptation to pathological voice detection. Notably, our system does not need target device sample labels, which allows for generalization to many new devices.

cs.LG↗

arXiv2018

The Effect of Heterogeneous Data for Alzheimer's Disease Detection from Speech

Speech datasets for identifying Alzheimer's disease (AD) are generally restricted to participants performing a single task, e.g. describing an image shown to them. As a result, models trained on linguistic features derived from such datasets may not be generalizable across tasks. Building on prior work demonstrating that same-task data of healthy participants helps improve AD detection on a single-task dataset of pathological speech, we augment an AD-specific dataset consisting of subjects describing a picture with multi-task healthy data. We demonstrate that normative data from multiple speech-based tasks helps improve AD detection by up to 9%. Visualization of decision boundaries reveals that models trained on a combination of structured picture descriptions and unstructured conversational speech have the least out-of-task error and show the most potential to generalize to multiple tasks. We analyze the impact of age of the added samples and if they affect fairness in classification. We also provide explanations for a possible inductive bias effect across tasks using model-agnostic feature anchors. This work highlights the need for heterogeneous datasets for encoding changes in multiple facets of cognition and for developing a task-independent AD detection model.

cs.LG↗

arXiv2018

ChainGAN: A sequential approach to GANs

We propose a new architecture and training methodology for generative adversarial networks. Current approaches attempt to learn the transformation from a noise sample to a generated data sample in one shot. Our proposed generator architecture, called $\textit{ChainGAN}$, uses a two-step process. It first attempts to transform a noise vector into a crude sample, similar to a traditional generator. Next, a chain of networks, called $\textit{editors}$, attempt to sequentially enhance this sample. We train each of these units independently, instead of with end-to-end backpropagation on the entire chain. Our model is robust, efficient, and flexible as we can apply it to various network architectures. We provide rationale for our choices and experimentally evaluate our model, achieving competitive results on several datasets.

cs.LG↗

arXiv2018

Semi-supervised classification by reaching consensus among modalities

Deep learning has demonstrated abilities to learn complex structures, but they can be restricted by available data. Recently, Consensus Networks (CNs) were proposed to alleviate data sparsity by utilizing features from multiple modalities, but they too have been limited by the size of labeled data. In this paper, we extend CN to Transductive Consensus Networks (TCNs), suitable for semi-supervised learning. In TCNs, different modalities of input are compressed into latent representations, which we encourage to become indistinguishable during iterative adversarial training. To understand TCNs two mechanisms, consensus and classification, we put forward its three variants in ablation studies on these mechanisms. To further investigate TCN models, we treat the latent representations as probability distributions and measure their similarities as the negative relative Jensen-Shannon divergences. We show that a consensus state beneficial for classification desires a stable but imperfect similarity between the representations. Overall, TCNs outperform or align with the best benchmark algorithms given 20 to 200 labeled samples on the Bank Marketing and the DementiaBank datasets.

cs.LG↗

arXiv2018

Augmenting word2vec with latent Dirichlet allocation within a clinical application

This paper presents three hybrid models that directly combine latent Dirichlet allocation and word embedding for distinguishing between speakers with and without Alzheimer's disease from transcripts of picture descriptions. Two of our models get F-scores over the current state-of-the-art using automatic methods on the DementiaBank dataset.

cs.CL↗

arXiv2018

Dropout during inference as a model for neurological degeneration in an image captioning network

We replicate a variation of the image captioning architecture by Vinyals et al. (2015), then introduce dropout during inference mode to simulate the effects of neurodegenerative diseases like Alzheimer's disease (AD) and Wernicke's aphasia (WA). We evaluate the effects of dropout on language production by measuring the KL-divergence of word frequency distributions and other linguistic metrics as dropout is added. We find that the generated sentences most closely approximate the word frequency distribution of the training corpus when using a moderate dropout of 0.4 during inference.

cs.CL↗

arXiv2017

On the importance of normative data in speech-based assessment

Data sets for identifying Alzheimer's disease (AD) are often relatively sparse, which limits their ability to train generalizable models. Here, we augment such a data set, DementiaBank, with each of two normative data sets, the Wisconsin Longitudinal Study and Talk2Me, each of which employs a speech-based picture-description assessment. Through minority class oversampling with ADASYN, we outperform state-of-the-art results in binary classification of people with and without AD in DementiaBank. This work highlights the effectiveness of combining sparse and difficult-to-acquire patient data with relatively large and easily accessible normative datasets.

cs.CL↗

arXiv2016

An Adaptive Psychoacoustic Model for Automatic Speech Recognition

Compared with automatic speech recognition (ASR), the human auditory system is more adept at handling noise-adverse situations, including environmental noise and channel distortion. To mimic this adeptness, auditory models have been widely incorporated in ASR systems to improve their robustness. This paper proposes a novel auditory model which incorporates psychoacoustics and otoacoustic emissions (OAEs) into ASR. In particular, we successfully implement the frequency-dependent property of psychoacoustic models and effectively improve resulting system performance. We also present a novel double-transform spectrum-analysis technique, which can qualitatively predict ASR performance for different noise types. Detailed theoretical analysis is provided to show the effectiveness of the proposed algorithm. Experiments are carried out on the AURORA2 database and show that the word recognition rate using our proposed feature extraction method is significantly increased over the baseline. Given models trained with clean speech, our proposed method achieves up to 85.39% word recognition accuracy on noisy data.

cs.CL↗

arXiv2016

Sex, drugs, and violence

Automatically detecting inappropriate content can be a difficult NLP task, requiring understanding context and innuendo, not just identifying specific keywords. Due to the large quantity of online user-generated content, automatic detection is becoming increasingly necessary. We take a largely unsupervised approach using a large corpus of narratives from a community-based self-publishing website and a small segment of crowd-sourced annotations. We explore topic modelling using latent Dirichlet allocation (and a variation), and use these to regress appropriateness ratings, effectively automating rating for suitability. The results suggest that certain topics inferred may be useful in detecting latent inappropriateness -- yielding recall up to 96% and low regression errors.

cs.CL↗

arXiv2016

Predicting health inspection results from online restaurant reviews

Informatics around public health are increasingly shifting from the professional to the public spheres. In this work, we apply linguistic analytics to restaurant reviews, from Yelp, in order to automatically predict official health inspection reports. We consider two types of feature sets, i.e., keyword detection and topic model features, and use these in several classification methods. Our empirical analysis shows that these extracted features can predict public health inspection reports with over 90% accuracy using simple support vector machines.

cs.CL↗

arXiv2015

Learning measures of semi-additive behaviour

In business analytics, measure values, such as sales numbers or volumes of cargo transported, are often summed along values of one or more corresponding categories, such as time or shipping container. However, not every measure should be added by default (e.g., one might more typically want a mean over the heights of a set of people); similarly, some measures should only be summed within certain constraints (e.g., population measures need not be summed over years). In systems such as Watson Analytics, the exact additive behaviour of a measure is often determined by a human expert. In this work, we propose a small set of features for this issue. We use these features in a case-based reasoning approach, where the system suggests an aggregation behaviour, with 86% accuracy in our collected dataset.

cs.AI↗

arXiv2010

Towards a Heuristic Categorization of Prepositional Phrases in English with WordNet

This document discusses an approach and its rudimentary realization towards automatic classification of PPs; the topic, that has not received as much attention in NLP as NPs and VPs. The approach is a rule-based heuristics outlined in several levels of our research. There are 7 semantic categories of PPs considered in this document that we are able to classify from an annotated corpus.

cs.CL↗