Offline Handwritten Text Recognition (HTR) is a task that offers a challenge in computer vision, where images are the only source of information. In fact, several approaches to optical models have been developed, such as through of Hidden Markov Model (HMM) or recurrent Bidirectional/Multidimensional layers. The current state-of-the-art consists of combined deep learning techniques, the Convolutional Recurrent Neural Networks (CRNN), in which recurrent layers still suffer from vanishing gradient problem when processing very long texts. In a way, high-performance models generally have millions of trainable parameters and a high computational cost. However, recently a new optical model architecture, Gated-CNN, demonstrated improvements to complement CRNN modeling. Thus, in this work, we present a new small architecture for HTR (based on Gated-CNN) integrated with two steps of language model at the character and word levels, respectively. Therefore, we used 9 state-of-the-art approaches and validated the results using the IAM public dataset. Finally, the proposed model surpasses the results obtained by different approaches in the literature, reaching recognition rates of CER 2.7% and WER 5.6%, which means an improvement of 13% over the best results on IAM dataset.

Huge amounts of musical manuscripts are preserved in cathedrals, abbeys, and archives. However, without reliable transcripts, their contents are inaccessible. Manual transcription is unaffordable for large collections, and current automatic technologies-such as Optical Music Recognition or Handwritten Music Recognition-do not provide sufficient accuracy for a fully-automatic scenario. In many cases, perfect transcripts are not really needed, given that content-based search with some degree of reliability would already be extremely useful. Spotting just single music symbols is rather useless (most of the symbols generally appear in all pages); instead, helpful search targets are melodic patterns, which typically correspond to music symbol sequences. We explore approaches for accurate retrieval of melodic patterns, represented by music symbol sequences, from collections of Medieval plainchant manuscripts. Our statistical framework, based on the use of convolutional recurrent neural networks and probabilistic indices, is shown to be useful for retrieving music patterns which appear frequently in this untranscribed images, yielding an Average Precision of 86 %.

Textual access to large collections of digitized images remains unfeasible because usually they lack transcripts. Transcribing such collections is in turn typically unattainable in terms of costs. However, the use of probabilistic indices can facilitate textual accessing with only moderate demands of resources. Besides allowing effortless information retrieval, it will be shown that probabilistic indices can also be used to estimate textual features of the indexed but otherwise untranscribed collections, such as running words and Zipf's curves. Complete probabilistic indices have been recently produced for two iconic large collections: "Bentham" (90K images) and "Spanish Golden Age Theater" (40K images). To show the repercussion of making these collections searchable, we provide accessing statistics gathered through their corresponding search interfaces. To the best of our knowledge this is the first publication of large collections of untranscribed manuscripts which are now publicly accessible for effective and efficient textual access.

State-of-the-art Handwritten Text Recognition techniques are based on statistical models such as hidden Markov models or recurrent neural networks for optical modeling of characters and N-grams for language modeling. These models are trained using well known, learning techniques: Expectation-Maximization, backpropagation, etc. Therefore, training data is needed to build these models. In the case of the optical models the training data consist of text line images with their corresponding transcripts. When the transcript of a handwritten document is available, putting in correspondence automatically the physical lines in the images with the lines of the transcripts is not an easy task. We present a method for automatically aligning handwritten text images and their respective transcripts. The approach automatically segments the images into lines and then recognizes them. An alignment confidence is obtained using the Levenshtein distance between the recognition results and the transcripts. The most confident lines are then used for training. Experiments carried out using a historical document present encouraging results.

Making large-scale collections of digitized historical documents searchable is being earnestly demanded by many archives and libraries. Probabilistically indexing the text images of these collections by means of keyword spotting techniques is currently seen as perhaps the only feasible approach to meet this demand. A vast medieval manuscript collection, written in both Latin and French, called "Chancery", is currently being considered for indexing at large. In addition to its bilingual nature, one of the major difficulties of this collection is the very high rate of abbreviated words which, on the other hand, are completely expanded in the ground truth transcripts available. In preparation to undertake full indexing of Chancery, experiments have been carried out on a relatively small but fully representative subset of this collection. To this end, a keyword spotting approach has been adopted which computes word relevance probabilities using character lattices produced by a recurrent neural network and a N-gram character language model. Results confirm the viability of the chosen approach for the large-scale indexing aimed at and show the ability of the proposed modeling and training approaches to properly deal with the abbreviation difficulties mentioned.

Music is a key element for cultural transmission, and so large collections of music manuscripts have been preserved over the centuries. In order to develop computational tools for analysis, indexing and retrieval from these sources, it is necessary to transcribe the content to some machine-readable format. In this paper we discuss the Handwritten Music Recognition problem, which refers to the development of automatic transcription systems for musical manuscripts. We focus on mensural notation, one of the most widespread varieties of Western classical music. For that, we present a labeled corpus containing 576 staves, along with a baseline recognition system based on a combination of hidden Markov models and N-gram language models. The baseline error obtained at symbol level is about 40 % which, given the difficulty of the task, can be considered a good starting point for future developments. Our aim is that these data and preliminary results help to promote this research field, serving as a reference in future developments.

This paper describes the fourth edition of the Handwritten Text Recognition (HTR) competition that was prepared this time in the context of the International Conference on Document Analysis and Recognition (ICDAR) 2017. Previous editions of this competition were conducted, first, with datasets from the tranScriptorium project in ICFHR 2014, and ICDAR 2015, and then, with datasets from the "Recognition and Enrichment of Archival Documents (READ)" European project in ICFHR 2016. This competition aims to bring together researchers working on off-line HTR and provides them a suitable benchmark to compare their techniques on the task of transcribing typical and difficult historical handwritten documents. The competition proposed for ICDAR 2017 aims at introducing a usual scenario for some collections in which there exist transcripts at page level for many pages useful for training, but these transcripts are not aligned with line images. Two tracks with different conditions on the use of training data were proposed. Most of the data comes from the Alfred Escher Letter Collection. But handwritten images were drawn from other German collections written by several hands.

Historical records of daily activities provide an intriguing look into the historic life. These documents have interesting information, useful for demography studies and genealogical research. However, automatic processing of historical documents, has mostly been focused on single works of literature and less on daily records, which tend to have a distinct layout, structure, and vocabulary. This paper presents a study about the capability of state-of-the-art handwritten text recognition and key word spotting systems, when applied to this kind of documents. A relatively small set of handwritten birth records registered in Wien in the 16th century is used in the experiments. A word accuracy of about 70% and an AP of 0.74 are achieved for plain image transcription and key word spotting respectively. Taking into account the many difficulties exhibited by these handwritten documents, these preliminary results are quite encouraging.

Traditionally, the HMM-Filler approach has been widely used in the fields of speech recognition and handwritten text recognition to tackle lexicon-free, query-by-string keyword spotting (KWS). It computes a score to determine whether a given keyword is written in a certain image region. It is conjectured, that this score is related to the confidence of the system, respect to the previous question. However, it is still not clear what this relationship is. In this paper, the HMM-Filler score is derived from a probabilistic formulation of KWS, which gives a better understanding of its behavior and limits. Additionally, the same probabilistic framework is used to present a new algorithm to compute the KWS scores, which results in better average precision (AP), for a keyword spotting task in the widely used IAM database. We show that the new algorithm can improve the HMM-filler results up to 10.4% relative (5.3% absolute) points in AP, in the considered task.

The principal goal of the Competition on Keyword Spotting for Handwritten Documents was to promote different approaches used in the field of Keyword Spotting and to fairly compare them using uniform data and metrics. To accommodate different perspectives adopted by researches in this field, the competition was divided into two distinct tracks, namely, a training-free and a training-based track, and each track entailed two optional assignments. Six participants submitted solutions to one or both assignments, depending on the capabilities and/or restrictions of their systems. The data used in the competition consisted of historical documents in English with different levels of complexity. This paper presents the details of the competition, including the data, evaluation metrics and results of the best participant methods.

This paper describes the second edition of the Handwritten Text Recognition (HTR) contest on the tranScriptorium datasets that has been held in the context of the International Conference on Document Analysis and Recognition 2015. Two tracks with different conditions on the use of training data were proposed. Nine research groups registered in the contest but finally three research submitted results. The handwritten images for this contest were drawn from the English “Bentham collection” dataset used in the tranScriptorium project. A small subset of this collection has been chosen for the present HTR competition. The selected subset has been written by several hands and entails significant variabilities and difficulties regarding the quality of text images, writing styles and crossed-out text. This contest is clearly more difficult than the the first edition both for training and for testing. A portion of the training dataset and the full test dataset were provided in the form of carefully segmented line images, along with the corresponding transcripts. Another portion of the training dataset was provided as raw images and their corresponding transcripts at region level. The three participants achieved good results, with transcription word error rates ranging from 31% down to 44%.

The so-called filler or garbage Hidden Markov Models (HMM-Filler) are among the most widely used models for lexicon-free, query by string key word spotting (KWS) in the fields of speech recognition and (lately) handwritten text recognition. However, it has important drawbacks. First, the keyword-specific HMM Viterbi decoding process needed to obtain the confidence scores of each spotted word involves a large computational cost. Second, in its traditional conception, the model does not take into account any context information - and more recent works where simple character bi-gram context is used show that not only the computational cost becomes even larger, but also the required keyword-specific language model becomes quite intricate to build. In a previous work we introduced KWS methods based on character lattices which proved very much simpler and faster than the traditional HMM-Filler, while providing practically identical results. Here we extend our previous work by using context-aware character lattices obtained by means of Viterbi decoding with high-order character N-gram models. Experimental results show that, as compared with a direct 2-gram HMM-filler implementation, the proposed approach requires between one and two orders of magnitude less query computing time. Moreover, for the first time in the field of handwritten text KWS, Filler-based results for N-grams up to N = 6 are reported, clearly showing a great impact of context on precision-recall performance.

Keyword Spotting (KWS) has been traditionally considered under two distinct frameworks: Query-by-Example (QbE) and Query-by-String (QbS). In both cases the user of the system wished to find occurrences of a particular keyword in a collection of document images. The difference is that, in QbE, the keyword is given as an exemplar image while, in QbS the keyword is given as a text string. In several works, the QbS scenario has been approached using QbE techniques; but the converse has not been studied in depth yet, despite of the fact that QbS systems typically achieve higher accuracy. In the present work, we present a very effective probabilistic approach to QbE KWS, based on highly accurate QbS KWS techniques which rely on models which need to be trained from annotated data. To assess the effectiveness of this approach, we tackle the segmentation-free QbE task of the ICFHR-2014 Competition on Handwritten KWS. Our approach achieves a mean average precision (mAP) as high as 0.715, which improves by more than 70% the best mAP achieved in this competition (0.419 under the same experimental conditions).

Tran Scriptorium is a 3-years project that aims to develop innovative, cost-effective solutions for the indexing, search and full transcription of historical handwritten document images, using Handwritten Text Recognition (HTR) technology. The production of ground-truth (GT) of a dataset of handwritten document images is among the first tasks. We address novel approaches for the faster production of this GT based on crowd-sourcing and on prior-knowledge methods. We also address here a novel low-cost semi-supervised procedure for obtaining pairs of correct line-level aligned detected/extracted text line images and text line transcripts, specially suitable for training models of the HTR technology employed in Tran Scriptorium.

Key-Word Spotting (KWS) in handwritten documents is approached here by means of Word Graphs (WG) obtained using segmentation-free handwritten text recognition technology based on N-gram Language Models and Hidden Markov Models. Linguistic context significantly boost KWS performance with respect to methods which ignore word contexts and/or rely on image-matching with pre-segmented isolated words. On the other hand, WG-based KWS can be significantly faster than other KWS approaches which directly work on the original images where, in general, computational demands are exceedingly high. A large WG contains most of the relevant information of the original text (line) image needed for KWS but, if it is too large, the computational advantages over traditional, image matching-based KWS become diminished. Conversely, if it is too small, relevant information may be lost, leading to degraded KWS precision/recall performance. We study the trade off between WG size and KWS information retrieval performance. Results show that small, computationally cheap WGs can be used without loosing the excellent KWS performance achieved with huge WGs.

The so-called filler or garbage Hidden Markov Models (HMM) are among the most widely used models for lexicon-free, query by string key word spotting in the fields of speech recognition and (lately) handwritten text recognition. An important drawback of this approach is the large computational cost of the keyword-specific HMM Viterbi decoding process needed to obtain the confidence scores of each word to be spotted. This paper presents a novel way to compute such confidence scores, directly from character lattices produced during a single Viterbi decoding process using only the "filler" model (i.e. no explicit keyword-specific decoding is needed). Experiments show that, as compared with the classical HMM-filler approach, the proposed method obtains essentially the same spotting results, while requiring between one and two orders of magnitude less query computing time.

The amount of digitized legacy documents has been rising dramatically over the last years due mainly to the increasing number of on-line digital libraries publishing this kind of documents, waiting to be classified and finally transcribed into a textual electronic format (such as ASCII or PDF). Nevertheless, most of the available fully-automatic applications addressing this task are far from being perfect and heavy and inefficient human intervention is often required to check and correct the results of such systems. In contrast, multimodal interactive-predictive approaches may allow the users to participate in the process helping the system to improve the overall performance. With this in mind, two sets of recent advances are introduced in this work: a novel interactive method for text block detection and two multimodal interactive handwritten text transcription systems which use active learning and interactive-predictive technologies in the recognition process.

To date, automatic handwriting recognition systems are far from being perfect and heavy human intervention is often required to check and correct the results of such systems. In order to achieve correct transcriptions, human knowledge can be integrated into the transcription process, following an Interactive Predictive paradigm. We have recently proposed Mouse Actions as a significant feedback information source for the underlying interactive system to improve the productivity of the human transcriptor. In this paper we review this way to interact with the system and report comparative results using the publicly available IAMDB dataset.

To date, automatic handwriting recognition systems are far from being perfect and often they need a post editing where a human intervention is required to check and correct the results of such systems. We propose to have a new interactive, on-line framework which, rather than full automation, aims at assisting the human in the proper recognition- transcription process; that is, facilitate and speed up their transcription task of handwritten texts. This framework combines the efficiency of automatic handwriting recognition systems with the accuracy of the human transcriptor. The best result is a cost-effective perfect transcription of the handwriting text images.

Pen-based interfaces aim at improving the man-machine interaction of many portable systems. While statistical models can be used to learn pen position sequences, they suffer from the huge variability exhibited by the speed of writing. To improve performance, invariance to the writing speed is needed. Trace segmentation is a technique that can be used to normalize the writing speed. This method is controlled by a parameter called resampling distance. A study of the resampling distance is presented here, along with another approximation to the writing speed normalization called "derivatives normalization". The improvement using trace segmentation was 193% relative to the baseline, whilst the improvement using derivatives normalization was 47.3% relative.