In this thesis, we address some of the challenges involved in developing a robust writer-independent, lexicon-free system to recognize online Tamil words. Tamil, being a Dravidian language, is morphologically rich and also agglutinative and thus does not have a finite lexicon. For example, a single verb root can easily lead to hundreds of words after morphological changes and agglutination. Further, adoption of a lexicon-free recognition approach can be applied to form-filling applications, wherein the lexicon can become cumbersome (if not impossible) to capture all possible names. Under such circumstances, one must necessarily explore the possibility of segmenting a Tamil word to its individual symbols. Modern day Tamil alphabet comprises 23 consonants and 11 vowels forming a total combination of 313 characters/aksharas. A minimal set of 155 distinct symbols have been derived to recognize these characters. A corpus of isolated Tamil symbols (IWFHR database) is used for deriving the various statistics proposed in this work. To address the challenges of segmentation and recognition (the primary focus of the thesis), Tamil words are collected using a custom application running on a tablet PC. A set of 10000 words (comprising 53246 symbols) have been collected from high school students and used for the experiments in this thesis. We refer to this database as the ‘MILE word database’. In the first part of the work, a feedback based word segmentation mechanism has been proposed. Initially, the Tamil word is segmented based on a bounding box overlap criterion. This dominant overlap criterion segmentation (DOCS) generates a set of candidate stroke groups. Thereafter, attention is paid to certain attributes from the resulting stroke groups for detecting any possible splits or under-segmentations. By relying on feedbacks provided by a priori knowledge of attributes such as number of dominant points and inter-stroke displacements the recognition label and likelihood of the primary SVM classifier linguistic knowledge on the detected stroke groups, a decision is taken to correct it or not. Accordingly, we call the proposed segmentation as ‘attention feedback segmentation’ (AFS). Across the words in the MILE word database, a segmentation rate of 99.7% is achieved at symbol level with AFS. The high segmentation rate (with feedback) in turn improves the symbol recognition rate of the primary SVM classifier from 83.9% (with DOCS alone) to 88.4%. For addressing the problem of segmentation, the SVM classifier fed with the x-y trace of the normalized and resampled online stroke groups is quite effective. However, the performance of the classifier is not robust to effectively distinguish between many sets of similar looking symbols. In order to improve the symbol recognition performance, we explore two approaches, namely reevaluation strategies and language models. The reevaluation techniques, in particular, resolve the ambiguities in base consonants, pure consonants and vowel modifiers to a considerable extent. For the frequently confused sets (derived from the confusion matrix), a dynamic time warping (DTW) approach is proposed to automatically extract their discriminative regions. Dedicated to each confusion set, novel localized cues are derived from the discriminative region for their disambiguation. The proposed features are quite promising in improving the symbol recognition performance of the confusion sets. Comparative experimental analysis of these features with x-y coordinates are performed for judging their discriminative power. The resolving of confusions is accomplished with expert networks, comprising discriminative region extractor, feature extractor and SVM. The proposed techniques improve the symbol recognition rate by 3.5% (from 88.4% to 91.9%) on the MILE word database over the primary SVM classifier. In the final part of the thesis, we integrate linguistic knowledge (derived from a text corpus) in the primary recognition system. The biclass, bigram and unigram language models at symbol level are compared in terms of recognition performance. Amongst the three models, the bigram model is shown to give the highest recognition accuracy. A class reduction approach for recognition is adopted by incorporating the language bigram model at the akshara level. Lastly, a judicious combination of reevaluation techniques with language models is proposed in this work. Overall, an improvement of up to 4.7% (from 88.4% to 93.1%) in symbol level accuracy is achieved. The writer-independent and lexicon-free segmentation-recognition approach developed in this thesis for online handwritten Tamil word recognition is promising. The best performance of 93.1% (achieved at symbol level) is comparable to the highest reported accuracy in the literature for Tamil symbols. However, the latter one is on a database of isolated symbols (IWFHR competition test dataset), whereas our accuracy is on a database of 10000 words and thus, a product of segmentation and classifier accuracies. The recognition performance obtained may be enhanced further by experimenting on and choosing the best set of features and classifiers. Also, the word recognition performance can be very significantly improved by using a lexicon. However, these are not the issues addressed by the thesis. We hope that the lexicon-free experiments reported in this work will serve as a benchmark for future efforts.

En esta tesis se presenta un nuevo marco interactivo y multimodal para la transcripción de Documentos manuscritos. Esta aproximación, lejos de proporcionar la transcripción completa pretende asistir al experto en la dura tarea de transcribir. Hasta la fecha, los sistemas de reconocimiento de texto manuscrito disponibles no proporcionan transcripciones aceptables por los usuarios y, generalmente, se requiere la intervención del humano para corregir las transcripciones obtenidas. Estos sistemas han demostrado ser realmente útiles en aplicaciones restringidas y con vocabularios limitados (como es el caso del reconocimiento de direcciones postales o de cantidades numéricas en cheques bancarios), consiguiendo en este tipo de tareas resultados aceptables. Sin embargo, cuando se trabaja con documentos manuscritos sin ningún tipo de restricción (como documentos manuscritos antiguos o texto espontáneo), la tecnología actual solo consigue resultados inaceptables. El escenario interactivo estudiado en esta tesis permite una solución más efectiva. En este escenario, el sistema de reconocimiento y el usuario cooperan para generar la transcripción final de la imagen de texto. El sistema utiliza la imagen de texto y una parte de la transcripción previamente validada (prefijo) para proponer una posible continuación. Despues, el usuario encuentra y corrige el siguente error producido por el sistema, generando así un nuevo prefijo mas largo. Este nuevo prefijo, es utilizado por el sistema para sugerir una nueva hipótesis. La tecnología utilizada se basa en modelos ocultos de Markov y n-gramas. Estos modelos son utilizados aquí de la misma manera que en el reconocimiento automático del habla. Algunas modificaciones en la definición convencional de los n-gramas han sido necesarias para tener en cuenta la retroalimentación del usuario en este sistema. / Romero Gómez, V. (2010). Multimodal Interactive Transcription of Handwritten Text Images [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/8541 / Palancia

This thesis aims to implement a segmentation-free strategy in the context of handwritten multi-digit string recognition. Three models namely VGG-16, CRNN and 4C are built to be evaluated and benchmarked, also research about the effect of the different training set on model performance is carried out.

Support vector machines (SVMs) are a powerful tool for classification problems. SVMs have only been developed in the last 20 years with the availability of cheap and abundant computing power. SVMs are a non-statistical approach and make no assumptions about the distribution of the data. Here support vector machines are applied to a classic data set from the machine learning literature and the out-of-sample misclassification rates are compared to other classification methods. Finally, an algorithm for using support vector machines to address the difficulty in imputing missing categorical data is proposed and its performance is demonstrated under three different scenarios using data from the 1997 National Labor Survey.

The design of a machine which reads unconstrained words still remains an unsolved problem. For example, automatic interpretation of handwritten documents by a computer is still under research. Most systems attempt to segment words into letters and read words one character at a time. However, segmenting handwritten words is very difficult. So to avoid this words are treated as a whole. This research investigates a number of features computed from whole words for the recognition of handwritten words in particular. Arabic text classification and recognition is a complicated process compared to Latin and Chinese text recognition systems. This is due to the nature cursiveness of Arabic text. The work presented in this thesis is proposed for word based recognition of handwritten Arabic scripts. This work is divided into three main stages to provide a recognition system. The first stage is the pre-processing, which applies efficient pre-processing methods which are essential for automatic recognition of handwritten documents. In this stage, techniques for detecting baseline and segmenting words in handwritten Arabic text are presented. Then connected components are extracted, and distances between different components are analyzed. The statistical distribution of these distances is then obtained to determine an optimal threshold for word segmentation. The second stage is feature extraction. This stage makes use of the normalized images to extract features that are essential in recognizing the images. Various method of feature extraction are implemented and examined. The third and final stage is the classification. Various classifiers are used for classification such as K nearest neighbour classifier (k-NN), neural network classifier (NN), Hidden Markov models (HMMs), and the Dynamic Bayesian Network (DBN). To test this concept, the particular pattern recognition problem studied is the classification of 32492 words using ii the IFN/ENIT database. The results were promising and very encouraging in terms of improved baseline detection and word segmentation for further recognition. Moreover, several feature subsets were examined and a best recognition performance of 81.5% is achieved.

No description available.

In this paper, we discuss the computer-aided processing of handwritten tabular records of historical weather data. The observationes meteorologicae, which are housed by the Regensburg University Library, are one of the oldest collections of weather data in Europe. Starting in 1771, meteorological data was consistently documented in a standardized form over almost 60 years by several writers. The tabular structure, as well as the unconstrained textual layout of comments and the use of historical characters, propose various challenges in layout and text recognition. We present a customized strategy to digitize tabular and handwritten data by combining various state-of-the-art methods for OCR processing to fit the collection. Since the recognition of historical documents still poses major challenges, we provide lessons learned from experimental testing during the first project stages. Our results show that deep learning methods can be used for text recognition and layout detection. However, they are less efficient for the recognition of tabular structures. Furthermore, a tailored approach had to be developed for the historical meteorological characters during the manual creation of ground truth data. The customized system achieved an accuracy rate of 82% for the text recognition of the heterogeneous handwriting and 87% accuracy for layout recognition of the tables.

Large variations in writing styles and difficulties in segmenting cursive words are the main reasons for handwritten cursive words recognition for being such a challenging task. An Indian postal document reading system based on a segmentation-free context based stochastic model is presented. The originality of the work resides on a combination of high-level perceptual features with the low-level pixel information considered by the former model and a pruning strategy in the Viterbi decoding to reduce the recognition time. While the low-level information can be easily extracted from the analyzed form, the discriminative power of such information has some limits as describes the shape with less precision. For that reason, we have considered in the framework of an analytical approach, using an implicit segmentation, the implant of high-level information reduced to a lower level. This enrichment can be perceived as a weight at pixel level, assigning an importance to each analyzed pixel based on their perceptual properties. The challenge is to combine the different type of features considering a certain dependence between them. To reduce the decoding time in the Viterbi search, a cumulative threshold mechanism is proposed in a flat lexicon representation. Instead of using a trie representation where the common prefix parts are shared we propose a threshold mechanism in the flat lexicon where based just on a partial Viterbi analysis, we can prune a model and stop the further processing. The cumulative thresholds are based on matching scores calculated at each letter level, allowing a certain dynamic and elasticity to the model. As we are interested in a complete postal address recognition system, we have also focused our attention on digit recognition, proposing different neural and stochastic solutions. To increase the accuracy and robustness of the classifiers a combination scheme is also proposed. The results obtained on different datasets written on Latin and Bengali scripts have shown the interest of the method and the recognition module developed will be integrated in a generic system for the Indian postal automation.

Esta pesquisa tenciona compreender as relações do escritor Mário de Andrade (1893-1945) com a literatura surrealista, considerando manuscritos e livros em seu acervo, no patrimônio do Instituto de Estudos Brasileiros da Universidade de São Paulo (IEB-USP), bem como a sua obra publicada. A pesquisa reúne e transcreve textos literários e ensaísticos atinentes ao Surrealismo, mencionados nos manuscritos do Fichário Analítico de Mário de Andrade; promove o levantamento de obras e de periódico surrealistas franceses presentes na biblioteca do criador de Macunaíma, recuperando anotações de leitura, as quais sugerem diálogos de criação na esfera da crítica literária e da literatura; congrega igualmente juízos críticos de Mário de Andrade sobre o Surrealismo expressos em sua obra édita e correspondência / This search intends to understand the relationship of the writer Mário de Andrade (1893 1945) with the Surrealist literature, considering handwritten and books in his collection, in the property of Institute of Brazilian Studies of São Paulo University (IEB-USP), as like his published work. The search gathers and transcripts literary and essayist texts pertinent to surrealism, mentioned in the handwritten of Mário de Andrades Analytical Card index; it organizes the survey of works and the French Surrealists periodic presents in the library of Macunaimas creator, recapturing reading notes, which suggest dialogues of creation in the literary critic sphere and the literature; it gathers equally Mário de Andrades criticizes judges about Surrealism expressed in his published work and correspondence.

Structured representation of input physical patterns as a set of local features has been useful for a veriety of robotics and human computer interaction (HCI) applications. It enables a stable understanding of the variable inputs. However, this representation does not fit the conventional machine learning algorithms and distance metrics because they assume vector inputs. To learn from input patterns with variable structure is thus challenging. To address this problem, I propose a general and systematic method to design distance metrics between structured inputs that can be used in conventional learning algorithms. Based on the observation of the stability in the geometric distributions of local features over the physical patterns across similar inputs, this is done combining the local similarities and the conformity of the geometric relationship between local features. The produced distance metrics, called “parametric kernels”, are positive semi-definite and require almost linear time to compute. To demonstrate the general applicability and the efficacy of this approach, I designed and applied parametric kernels to handwritten character recognition, on-line face recognition, and object detection from laser range finder sensor data. Parametric kernels achieve recognition rates competitive to state-of-the-art approaches in these tasks. / text