Les méthodes à l’état de l’art de la reconnaissance de l’écriture sont fondées sur des réseaux de neurones récurrents (RNN) à cellules LSTM ayant des performances remarquables. Dans cette thèse, nous proposons deux nouveaux principes la vérification lexicale et la génération de cohorte afin d’attaquer les problèmes de la reconnaissance de l’écriture : i) le problème des grands lexiques et des décodages dirigés par le lexique ii) la problématique de combinaison de modèles optiques pour une meilleure reconnaissance iii) la nécessité de constituer de très grands ensembles de données étiquetées dans un contexte d’apprentissage profond. La vérification lexicale est une alternative aux décodages dirigés par le lexique peu étudiée à cause des faibles performances des modèles optiques historiques (HMM). Nous montrons dans cette thèse qu’elle constitue une alternative intéressante aux approches dirigées par le lexique lorsqu’elles s’appuient sur des modèles optiques très performants comme les RNN LSTM. La génération de cohorte permet de générer facilement et rapidement un grand nombre de réseaux récurrents complémentaires en un seul apprentissage. De ces deux techniques nous construisons et proposons un nouveau schéma de cascade pour la reconnaissance de mots isolés, une nouvelle combinaison au niveau ligne LV-ROVER et une nouvelle stratégie d’auto-apprentissage de RNN LSTM pour la reconnaissance de mots isolés. La cascade proposée permet de combiner avec la vérification lexicale des milliers de réseaux et atteint des résultats à l’état de l’art pour les bases Rimes et IAM. LV-ROVER a une complexité réduite par rapport à l’algorithme original ROVER et permet de combiner des centaines de réseaux sans modèle de langage tout en dépassant l’état de l’art pour la reconnaissance de lignes sur le jeu de donnéesRimes. Notre stratégie d’auto-apprentissage permet d’apprendre à partir d’un seul réseau BLSTM et sans paramètres grâce à la cohorte et la vérification lexicale, elle montre d’excellents résultats sur les bases Rimes et IAM. / State-of-the-art methods for handwriting recognition are based on LSTM recurrent neural networks (RNN) which achieve high performance recognition. In this thesis, we propose the lexicon verification and the cohort generation as two new building blocs to tackle the problem of handwriting recognition which are : i) the large vocabulary problem and the use of lexicon driven methods ii) the combination of multiple optical models iii) the need for large labeled dataset for training RNN. The lexicon verification is an alternative to the lexicon driven decoding process and can deal with lexicons of 3 millions words. The cohort generation is a method to get easily and quickly a large number of complementary recurrent neural networks extracted from a single training. From these two new techniques we build and propose a new cascade scheme for isolated word recognition, a new line level combination LV-ROVER and a new self-training strategy to train LSTM RNN for isolated handwritten words recognition. The proposed cascade combines thousands of LSTM RNN with lexicon verification and achieves state-of-the art word recognition performance on the Rimes and IAM datasets. The Lexicon Verified ROVER : LV-ROVER, has a reduce complexity compare to the original ROVER algorithm and combine hundreds of recognizers without language models while achieving state of the art for handwritten line text on the RIMES dataset. Our self-training strategy use both labeled and unlabeled data with the unlabeled data being self-labeled by its own lexicon verified predictions. The strategy enables self-training with a single BLSTM and show excellent results on the Rimes and Iam datasets.

The purpose of this study is to investigate whether there are differences between students' narrative texts written by hand or with an iPad (with keyboard). This comparative study analyzes 54 narrative texts written by students in grade 6. The method of analysis is a combination of a qualitative text analysis model with a quantitative content analysis of the texts, with different measurable variables. The text analysis model systematizes the content of the texts in the categories content, structure and linguistic features, and writing rules. The quantitative analysis is based on measurable variables within the framework of the categories and these are text length, word class distribution, nominal quota, paragraph classification, word length and word variation, readability index, misspellings and spelling and verbal words. The quantitative study has been processed statistically with Swegram. The results show that there are differences in the student grades written by hand and digitally in all three categories. The digital texts are longer, contain more descriptive descriptions, a clearer structure, a more varied language and less spelling errors. But the entirety of the digital texts is not exceptional. Students use a more complicated writing style, vary their language and use more difficult words. However, many students do not have the knowledge, skills or the technical writing ability to compose a clarity of communicate the content in the text. Writing digitally has great advantages over handwriting, but there are also disadvantages. / Syftet med denna studie är att undersöka om det finns skillnader i elevers berättande texter när de skriver för hand med papper och penna eller skriver på iPad med tangentbord (digi­talt verktyg). I denna jämförande studie analyseras 54 berättande texter skrivna av elever i årskurs 6. Analysmetoden är en kombination av en kvalitativ textanalysmodell och en kvantitativ innehållsanalys av texterna, med olika mätbara variabler. Textanalysmodellen systematiserar texternas innehåll i kategorierna innehåll, struktur samt språkliga drag och skrivregler. Den kvantitativa analysen bygger på mätbara variabler inom ramen för kate­gorierna. Dessa är textlängd, ordklassfördelning, nominalkvot, styckeindelning, ordlängd och ordvariation, läsbarhetsindex, stavfel och särskrivningar och talspråk. Den kvantita­tiva studien har bearbetats statistiskt med hjälp av Swegram. Resultatet visar att det finns skillnader i elevtexterna skrivna för hand och digitalt inom alla tre kategorierna. De digitala texterna är längre, innehåller fler gestaltande beskrivningar och har tydligare struktur, ett mer varierat språk och färre stavfel. Men de digitala texternas helhet är inte alltid bättre. Eleverna använder ett mer avancerat skrivsätt, varierar sitt språk och använder svårare ord. Emellertid har flera elever inte kunskapen eller den skrivtekniska förmågan att komponera tydlighet i budskap och innehåll. Att skriva digitalt har stora fördelar framför att skriva för hand men det finns även nackdelar.

Der Teilnachlass des prominenten DDR-Chefarchitekten Hermann Henselmann (1905-1995) in der Handschriftensammlung der Sächsischen Landesbibliothek – Staats- und Universitätsbibliothek Dresden steht für ein sehr junges Kapitel deutscher Architekturgeschichte. ...

Denna undersökning handlar om handskriftens närvaro i den digitala skolan 2021 och vilka attityder lärare och elever har till att skriva för hand respektive att skriva på datorn samt fördelar och nackdelar med de båda skrivmetoderna. Empirin har samlats in genom observationer i klassrum och intervjuer med lärare och elever. Samtalen var digitala för ena gruppen och genom enskilda fysiska möten i den andra gruppen. Resultaten i denna undersökning gav både väntade och oväntade svar. Eleverna ansåg att de skrev längst texter på datorn tack vare rättstavningsprogrammen vilket var ett förväntat svar på frågan. Men att två tredjedelar av eleverna ansåg att de skrev en text snabbare för hand än på datorn var en oväntad vinkling. Lärarnas attityder handlade om skriftsättens olika fördelar och nackdelar. Empirin gav förväntade svar om hur bra handskrift är för det långsiktiga lärandet och de digitala hjälpmedlens möjligheter. Observationen visade att det skrivs både för hand och på datorn i skolan idag samt att skillnaden i tid mellan skrivsätten är mindre än jag hade uppfattningen om när jag började samla in empirin. / This survey is about the handwriting's presence in the digital school in 2021 and what attitudes teachers and students have to writing by hand and writing on the computer, respectively and advantages and disadvantages of both writing methods. The empirics have been collected through classroom observations and interviews with teachers and students. The conversations were digital for one group and through individual physical meetings in the other group. The results in this examination provided both expected and unexpected answers. The students felt that they wrote the longest texts on the computer thanks to the spelling programs, which was an expected answer to the question. But the fact that two-thirds of students thought they were writing a text faster by hand than on the computer was an unexpected slant. Teachers' attitudes were about the different advantages and disadvantages of the two types of writing. Empirical services provided expected answers about the quality of handwriting for long-term learning and the possibilities of digital aids. The observation showed that it is written both by hand and on the computer at school today and that the difference in time between writing methods is smaller than I had the idea of when I started collecting the empirical.

Syftet med studien är att genom en kvalitativ metod bidra till befintlig forskning om olika strategier som elever i årskurs 1 och 2 använder sig av vid handskrivning. I studien använder jag mig av individuella intervjuer med både lärare och elever, samt observationer av elever när de skriver med penna. Fenomenografi har använts som metod vid databearbetningen. Genom en kvalitativ metod och ”embodied cognition” som teoretisk utgångspunkt får jag både höra vilka strategier eleverna beskriver att de använder sig av, men även se vilka kroppsliga rörelser och strategier de uppvisar. Det framkommer även hur lärarna beskriver sin skrivundervisning och att de inte har tid att notera elevernas kroppsliga uttryck. Resultatet åskådliggör att eleverna uppvisar många kroppsliga rörelser vid handskrivning. Med stöd i tidigare forskning visar sig dessa rörelser vara både kognitiva och kroppsliga strategier, vilka hjälper eleverna i deras skrivprocess. Ibland signalerar även rörelserna behov av hjälp, trots att eleverna inte själva är medvetna om dem. / The purpose of this study is to, through qualitative methods, contribute to existing research about different strategies that pupils in grades 1 and 2 make use of in handwriting. In the study, I carry out individual interviews with both teachers and pupils, along with observations of pupils when they are writing with pencils. The method used in data processing is phenomenography. Through a qualitative method and “embodied cognition” as a theoretical starting point, the pupils describe which strategies that they use, but also which body movements and strategies they show. The result also shows how the teachers describe their writing instruction and that they do not have time to take note of the pupils’ bodily expressions. The result illustrates that the pupils display many bodily movements during handwriting. Drawing on earlier research, this study shows that these movements are both cognitive and bodily strategies, which help pupils in their writing process. Sometimes the movements signal a need of help, even though the pupils are not aware of it themselves.

I den digitala utveckling som sker inom skolans värld använder eleverna datorn i allt större utsträckning när de producerar texter. Som en del i att skapa en vetenskaplig grund för att förstå hur digitala verktyg kan påverka textskapande är syftet med denna studie att undersöka på vilket sätt det medierande redskapet för att producera text, i detta fall tangentbord, skärm och ordbehandlingsprogram eller papper och penna, förändrar texters transitivitet och hur idéationella betydelser realiseras i de olika texttyperna. Studien ämnar även undersöka hur agentiviteten i texter förändras beroende på medierande redskap och vilka skillnader som det går att identifiera i nyttjandet av olika pronomen i agentiva processer. Tjugo texter skrivna av elever i årskurs 4 och 5 analyserades med utgångspunkt i systemisk funktionell textanalys och analysmetoderna transitivitetsanalys och ergativitetsanalys. Studiens resultat är inkonklusiva och ingen slutsats kan dras om skrivverktygens påverkan på texternas transitivitet eller agentivitet. Studien visar hur uppgiftbeskrivningar för elevers skolskrivande är mer avgörande för hur innehållet konstrueras än om de skriver för hand eller på datorn och bidrar med kunskap för vidare forskning. / In today’s developing digital era, students in school rely to an increasing extent when producing texts. As part of an effort to contribute to creating a scientific basis in regard to understanding how digital tools affects how text is constructed, this study aims to explore in which way the mediating tool for text production, in this case keyboard, screen, and word processing software versus paper and pen, changes the transitivity in texts and how ideational meaning is realised. Additionally, the study investigates how the ergativity in texts change depending on mediating tool and what differences in the use of pronouns can be identified in ergativ processes. Twenty texts written by fourth and fifth grade students are analysed based on Systemic-Functional text analysis and the analysis methods of transitivity analysis and ergativity analysis. The study’s results are inconclusive, providing no clear conclusions on the impact of writing tools on the transitivity or agency of texts. However, the study reveals that task descriptions for students' writing are more critical in shaping content than the choice between handwriting or typing, contributing valuable insights for further research.

No description available.

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.

No description available.