Salvador España-Boquera

Fast Evaluation of Connectionist Language Models

Connectionist language models offer many advantages over their statistical counterparts, but they also have some drawbacks like a much more expensive computational cost. This paper describes a novel method to overcome this problem. A set of normalization values associated to the most frequent n -gramsis pre-computed and the model is smoothed with lower n -gramconnectionist or statistical models. The proposed approach is favourably compared to standard connectionist language models and with statistical back-off language models.

F-measure as the error function to train neural networks

Imbalance datasets impose serious problems in machine learning. For many tasks characterized by imbalanced data, the F-Measure seems more appropiate than the Mean Square Error or other error measures. This paper studies the use of F-Measure as the training criterion for Neural Networks by integrating it in the Error-Backpropagation algorithm. This novel training criterion has been validated empirically on a real task for which F-Measure is typically applied to evaluate the quality. The task consists in cleaning and enhancing ancient document images which is performed, in this work, by means of neural filters.

Improving a DTW-Based Recognition Engine for On-line Handwritten Characters by Using MLPs

Our open source real-time recognition engine for on-line isolated handwritten characters is a 3-Nearest Neighbor classifier that uses approximate dynamic time warping comparisons with a set of prototypes filtered by two fast distance-based methods. This engine achieved excellent classification rates on two writer-independent tasks:UJIpenchars and Pendigits. We present the integration of multilayer perceptrons into our engine, an improvement that speeds up the recognition process by taking advantage of the independence of these networks’ classification times from training set sizes. We also present experimental results on our new publicly available UJIpenchars2 database and on Pendigits.

Insights on the Use of Convolutional Neural Networks for Document Image Binarization

Convolutional Neural Networks have systematically shown good performance in Computer Vision and in Handwritten Text Recognition tasks. This paper proposes the use of these models for document image binarization. The main idea is to classify each pixel of the image into foreground and background from a sliding window centered at the pixel to be classified. An experimental analysis on the effect of sensitive parameters and some working topologies are proposed using two different corpora, of very different properties: DIBCO and Santgall.

Cache neural network language models based on long-distance dependencies for a spoken dialog system

The integration of a cache memory into a connectionist language model is proposed in this paper. The model captures long term dependencies of both words and concepts and is particularly useful for Spoken Language Understanding tasks. Experiments conducted on a human-machine telephone dialog corpus are reported, and an increase in performance is observed when features of previous turns are taken into account for predicting the concepts expressed in a user turn. In terms of Concept Error Rate we obtained a statistically significant improvement of 3.2 points over our baseline (10% relative improvement) on the French Media corpus.

Unconstrained offline handwriting recognition using connectionist character N-grams

This work presents an unconstrained offline hand-written line recognition system based on hybrid HMM (Hidden Markov Model)/ANN (Artificial Neural Network) models. The particularity of the system lies in the use of an ensemble of connectionist/statistical character n-gram language models. These language models are trained with a text corpus at character level; therefore, no explicit lexicon is used during recognition. The recognizer is thus able to output words which do not belong to that corpus. The proposed system favorably behaves compared to using a standard character n-gram on the IAM database lines corpus and achieves error rates comparable to state-of-the-art lexicon-driven alternatives.

Improving isolated handwritten word recognition using a specialized classifier for short words

The aim of this work is to improve the performance of off-line handwritten text recognition systems based on hidden Markov models (HMM) and hybrid Markov models with neural networks (HMM/ANN). In order to study the systems without the influence of the language model, an isolated word recognition task has been performed. The analysis of the influence of word lengths on the error rates of the recognizers has lead to combine those classifiers with another one specialized in short words. To this end, various multilayer perceptrons have been trained to classify a subset of the vocabulary in a holistic manner. Combining the classifiers by means of a variation of the Borda count voting method achieves very satisfying results.

A flexible system for document processing and text transcription

STATE is a flexible system for document processing. It comprises a graphical front-end that can be easily connected to different text recognition back-ends. We comment here the front-end and two backends: one based on nearest neighbors and one based on Hidden Markov Models. Experimentation shows that if the back-end has a moderately low character error rate, productivity gains can be as high as 100% when compared to directly transcribing the text.

Speech interaction in a multimodal tool for handwritten text transcription

STATE is a multimodal tool for document processing and text transcription. Its graphical front-end can be easily connected to different text recognition back-ends. New features and improvements are presented in this work: the interactive correction of one word in the transcribed line has been improved to reestimate the entire transcription line using the user feedback and speech input has been integrated in the multimodal interface enabling the user to also utter the word to be corrected, giving the user the possibility to use the interface according to her preferences or the task at hand. Thus, at the current version of STATE, the user can type, write on the screen with a stylus, or utter the incorrectly recognized word, and then, the system uses the user feedback in any of the proposed modalities to reestimate the transcribed line so as to hopefully correct other errors which could be caused by the mistaken word the user has corrected.

Behaviour-based clustering of neural networks applied to document enhancement

This work proposes an agglomerative hierarchical clustering algorithm where the items to be clustered are supervised-learning classifiers. The measure of similarity to compare classifiers is based on their behaviour. This clustering algorithm has been applied to document enhancement: A set of neural filters is trained with multilayer perceptrons for different types of noise and then clustered into groups to obtain a reduced set of neural clustered filters. In order to automatically determine which clustered filter is the most suitable to clean and enhance a real noisy image, an image classifier is also trained using multilayer perceptrons.