A5073107684- Natural Language Processing Techniques
- Handwritten Text Recognition Techniques
- Topic Modeling
- Algorithms and Data Compression
- Image Processing and 3D Reconstruction
- Machine Learning and Algorithms
- Image Retrieval and Classification Techniques
- Music and Audio Processing
- Speech Recognition and Synthesis
- Speech and dialogue systems
- Advanced Image and Video Retrieval Techniques
- Face and Expression Recognition
- Data Management and Algorithms
- Hand Gesture Recognition Systems
- Machine Learning and Data Classification
- semigroups and automata theory
- Vehicle License Plate Recognition
- Time Series Analysis and Forecasting
- Text and Document Classification Technologies
- Speech and Audio Processing
- Neural Networks and Applications
- Video Analysis and Summarization
- Iterative Learning Control Systems
- Music Technology and Sound Studies
- Web Data Mining and Analysis
Universitat Politècnica de València
2015-2024
Exploratorium
2024
Artificial Intelligence Research Institute
2023
Language Science (South Korea)
2022
Universitat Politècnica de Catalunya
2021
Consejo de Ciencia y Tecnología del Estado de Tabasco
2018
Universidad de Lima
2016
Centro Tecnológico de Investigación, Desarrollo e Innovación en tecnologías de la Información y las Comunicaciones (TIC)
2003-2012
Universitat Jaume I
1998-2008
RWTH Aachen University
2008
Given two strings X and Y over a finite alphabet, the normalized edit distance between Y, d(X,Y) is defined as minimum of W(P)/L(P), where P an editing path W(P) sum weights elementary operations P, L(P) number these (length P). It shown that in general, cannot be computed by first obtaining conventional (unnormalized) then normalizing this value length corresponding path. In order to compute distances, algorithm can implemented work O(m*n/sup 2/) time O(n/sup memory space proposed, m n are...
Probabilistic finite-state machines are used today in a variety of areas pattern recognition or fields to which is linked. In Part I this paper, we surveyed these objects and studied their properties. II, study the relations between probabilistic automata other well-known devices that generate strings like hidden Markov models n-grams provide theorems, algorithms, properties represent current state art objects.
Current machine translation (MT) systems are still not perfect. In practice, the output from these needs to be edited correct errors. A way of increasing productivity whole process (MT plus human work) is incorporate correction activities within itself, thereby shifting MT paradigm that computer-assisted translation. This model entails an iterative in which translator activity included loop: each iteration, a prefix validated (accepted or amended) by and system computes its best (or n-best)...
The inductive inference of the class k-testable languages in strict sense (k-TSSL) is considered. A k-TSSL essentially defined by a finite set substrings length k that are permitted to appear strings language. Given positive sample R an unknown language, deterministic finite-state automation recognizes smallest containing obtained. inferred shown have number transitions bounded O(m) where m defining this k-TSSL, and algorithm works O(kn log m) n sum lengths all R. proposed methods...
A formalization of the transducer learning problem and an effective efficient method for inductive important class transducers, subsequential are presented. The capabilities transductions illustrated through a series experiments that also show high effectiveness proposed in obtaining very accurate compact transducers corresponding tasks.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
In order to optimize the accuracy of nearest-neighbor classification rule, a weighted distance is proposed, along with algorithms automatically learn corresponding weights. These weights may be specific for each class and feature, individual prototype, or both. The learning are derived by (approximately) minimizing leaving-one-out error given training set. proposed approach assessed through series experiments UCI/STATLOG corpora, as well more task text which entails very sparse data...
This paper describes the Handwritten Text Recognition (HTR) competition on READ dataset that has been held in context of International Conference Frontiers Handwriting 2016. aims to bring together researchers working off-line HTR and provide them a suitable benchmark compare their techniques task transcribing typical historical handwritten documents. Two tracks with different conditions use training data were proposed. Ten research groups registered but finally five submitted results. The...
Purpose An overview of the current use handwritten text recognition (HTR) on archival manuscript material, as provided by EU H2020 funded Transkribus platform. It explains HTR, demonstrates , gives examples cases, highlights affect HTR may have scholarship, and evidences this turning point advanced digitised heritage content. The paper aims to discuss these issues. Design/methodology/approach This adopts a case study approach, using development delivery one openly available platform for...
Finite-state transducers are models that being used in different areas of pattern recognition and computational linguistics. One these is machine translation, which the approaches based on building automatically from training examples becoming more attractive. very adequate for use constrained tasks samples pairs sentences available. A technique inferring finite-state proposed this article. This formal relations between rational grammars. Given a corpus source-target sentences, approach uses...
A fully integrated approach to speech input language translation in limited domain applications is presented. The mapping from the output modeled terms of a finite state model which learned examples sentences task considered. This tightly with standard acoustic phonetic models and resulting global directly supplies, through Viterbi search, an optimal sentence for each utterance. Several extensions this framework, recently developed cope increasing difficulty tasks, are reviewed. Finally,...
The interpretation of handwritten sentences is carried out using a holistic approach in which both text image recognition and the itself are tightly integrated. Conventional approaches follow serial, first-recognition then-interpretation scheme cannot adequately use semantic–pragmatic knowledge to recover from errors. Stochastic finite-sate transducers shown be suitable models for this integration, permitting full exploitation final constraints. Continuous-density hidden Markov embedded...
A contest on Handwritten Text Recognition organised in the context of ICFHR 2014 conference is described. Two tracks with increased freedom use training data were proposed and three research groups participated these two tracks. The handwritten images for this drawn from an English set which currently being considered Tran scriptorium project. goal project to develop innovative, efficient cost-effective solutions transcription historical document images, focusing four languages: English,...
This work details the results of a face authentication test (FAT2004) (http://www.ee.surrey.ac.uk/banca/icpr2004) held in conjunction with 17th International Conference on Pattern Recognition. The contest was publicly available BANCA database (http://www.ee.surrey.ac.uk/banca) according to defined protocol (E. Bailly-Bailliere et al., June 2003). competition also had sequestered part which institutions submit their algorithms for independent testing. 13 different verification from 10...
This paper describes the fourth edition of Handwritten Text Recognition (HTR) competition that was prepared this time in context International Conference on Document Analysis and (ICDAR) 2017. Previous editions were conducted, first, with datasets from tranScriptorium project ICFHR 2014, ICDAR 2015, then, "Recognition Enrichment Archival Documents (READ)" European 2016. aims to bring together researchers working off-line HTR provides them a suitable benchmark compare their techniques task...