Ciphers are encrypted documents created to hide their content from those who were not the receivers of the message. Different types of symbols, such as zodiac signs, alchemical symbols, alphabet letters or digits are exploited to compose the encrypted text which needs to be decrypted to gain access to the content of the documents. The first step before decryption is the transcription of the cipher. The purpose of this thesis is to evaluate an automatic transcription tool from image to a text format to provide a transcription of the cipher images. We implement a supervised few-shot deep-learning model which is tested on different types of encrypted documents and use various evaluation metrics to assess the results. We show that the few-shot model presents promising results on seen data with Symbol Error Rates (SER) ranging from 8.21% to 47.55% and accuracy scores from 80.13% to 90.27%, whereas SER in out-of-domain datasets reaches 79.91%. While a wide range of symbols are correctly transcribed, the erroneous symbols mainly contain diacritics or are punctuation marks.

Thousands of historical ciphers, encrypted manuscripts, are stored in archives across Europe. Historical cryptology is the research field concerned with studying these manuscripts - combining the interest of humanistic fields with methods of cryptography and computational linguistics. Before a cipher can be decrypted by automatic means, it must first be transcribed into machine-readable digital text. Image processing techniques and Deep Learning have enabled transcription of handwritten text to be performed automatically, but the task faces challenges when ciphers constitute the target data. The main reason is a lack of labeled data, caused by the heterogeneity of handwriting and the tendency of ciphers to employ unique symbol sets. Few-Shot Learning is a machine learning framework which reduces the need for labeled data, using pretrained models in combination with support sets containing a few labeled examples from the target data set. This project is concerned with evaluating a Few-Shot model on the task of transcription of historical ciphers. The model is tested on pages from three in-domain ciphers which vary in handwriting style and symbol sets. The project also investigates the use of further fine-tuning the model by training it on a limited amount of labeled symbol examples from the respective target ciphers. We find that the performance of the model is dependant on the handwriting style of the target document, and that certain model parameters should be explored individually for each data set. We further show that fine-tuning the model is indeed efficient, lowering the Symbol Error Rate (SER) at best 27.6 percentage points.

Digital libraries and archives are major portals to rich sources of information. They undertake large-scale digitization to enhance their digital collections and offer users valuable text data. When it comes to handwritten documents, usually these are only provided as digitized images and not accompanied by their transcriptions. Text in non-machine-readable format restricts contemporary scholars to conduct research, especially by employing digital humanities approaches, such as distant reading and data mining. The purpose of this thesis is to evaluate Transkribus platform as a linguistic tool mainly developed for producing automatic transcriptions of handwritten documents. The results are correlated with the findings of a questionnaire distributed to libraries and archives across Europe to expand our knowledge on the policy they follow regarding manuscripts and transcription provision. A model for a specific writing style in Latin language is trained and the accuracy on various Latin handwritten pages is tested. Finally, the tool’s validation is discussed, as well as to what extent it meets the general needs of the cultural heritage institutions and of humanities scholars.

En traitement du signal audio, l’analyse est une étape essentielle permettant de comprendre et d’inter-agir avec les signaux existants. En effet, la qualité des signaux obtenus par transformation ou par synthèse des paramètres estimés dépend de la précision des estimateurs utilisés. Cependant, des limitations théoriques existent et démontrent que la qualité maximale pouvant être atteinte avec une approche classique peut s’avérer insuffisante dans les applications les plus exigeantes (e.g. écoute active de la musique). Le travail présenté dans cette thèse revisite certains problèmes d’analyse usuels tels que l’analyse spectrale, la transcription automatique et la séparation de sources en utilisant une approche dite “informée”. Cette nouvelle approche exploite la configuration des studios de musique actuels qui maitrisent la chaîne de traitement avant l’étape de création du mélange. Dans les solutions proposées, de l’information complémentaire minimale calculée est transmise en même temps que le signal de mélange afin de permettre certaines transformations sur celui-ci tout en garantissant le niveau de qualité. Lorsqu’une compatibilité avec les formats audio existants est nécessaire, cette information est cachée à l’intérieur du mélange lui-même de manière inaudible grâce au tatouage audionumérique. Ce travail de thèse présente de nombreux aspects théoriques et pratiques dans lesquels nous montrons que la combinaison d’un estimateur avec de l’information complémentaire permet d’améliorer les performances des approches usuelles telles que l’estimation non informée ou le codage pur. / In the field of audio signal processing, analysis is an essential step which allows interactions with existing signals. In fact, the quality of transformed or synthesized audio signals depends on the accuracy over the estimated model parameters. However, theoretical limits exist and show that the best accuracy which can be reached by a classic estimator can be insufficient for the most demanding applications (e.g. active listening of music). The work which is developed in this thesis revisits well known audio analysis problems like spectral analysis, automatic transcription of music and audio sources separation using the novel ``informed'' approach. This approach takes advantage of a specific configuration where the parameters of the elementary signals which compose a mixture are known before the mixing process. Using the tools which are proposed in this thesis, the minimal side information is computed and transmitted with the mixture signal. This allows any kind of transformation of the mixture signal with a constraint over the resulting quality. When the compatibility with existing audio formats is required, the side information is embedded directly into the analyzed audio signal using a watermarking technique. This work describes several theoretical and practical aspects of audio signal processing. We show that a classic estimator combined with the sufficient side information can obtain better performances than classic approaches (classic estimation or pure coding).

Este trabalho relaciona elementos das áreas de Computação e Saúde para comporem a elaboração de uma metodologia para avaliação de usabilidade de sistemas de transcrição automática de laudos na área de Radiologia. Inicialmente, é apresentado um estudo realizado sobre a área de Interface do Usuário Baseada em Voz que identifica requisitos para os sistemas que trabalham com comunicação mediada por voz assim como as iniciativas no sentido de se criar uma metodologia para sua avaliação. Em seguida é realizado um estudo dos sistemas de transcrição automática de laudos, no qual os principais requisitos são caracterizados e classificados. Os dois estudos acima citados foram integrados para a elaboração de uma metodologia para a avaliação de Sistemas de Transcrição Automática de Laudos em Radiologia. A metodologia foi, então, validada previamente através de inspeções e testes de usabilidade realizados fora do ambiente hospitalar, através do uso de um sistema de transcrição automática de laudos em Radiologia. Posteriormente, a metodologia foi também aplicada a um hospital da cidade de São Paulo. Como resultado principal foi proposto um guia bastante detalhado para se avaliar os Sistemas de Transcrição Automática de Laudos em Radiologia, cada vez mais presentes em hospitais e clínicas no país, além dos relatos das experiências obtidas com a aplicação desta metodologia em um caso real. / This work combines knowledge from Computer Science and Health Science in order to propose an evaluation methodology for Automatic Transcription System of Radiology Reports. At first a study regarding a Voice User Interface is presented, this interface identifies the requirements for Spoken Language Dialogue Systems and it can also be used as a tool for an evaluation methodology. Following, a study of automatic transcription systems is presented; in this study the main requirements are listed and classified. Both studies were integrated to allow a new methodology to evaluate an Automatic Transcription System in Radiology. This methodology was previously validated through some inspections and usability tests outside the hospital environment and afterword the methodology was used in a hospital in São Paulo city. As a main result it was proposed a very detailed guide for evaluating the Automatic Transcription Systems Reports in Radiology, increasingly found in hospitals and clinics in this country, apart from reports on experiences gained in applying this methodology in a real case.

Este trabalho relaciona elementos das áreas de Computação e Saúde para comporem a elaboração de uma metodologia para avaliação de usabilidade de sistemas de transcrição automática de laudos na área de Radiologia. Inicialmente, é apresentado um estudo realizado sobre a área de Interface do Usuário Baseada em Voz que identifica requisitos para os sistemas que trabalham com comunicação mediada por voz assim como as iniciativas no sentido de se criar uma metodologia para sua avaliação. Em seguida é realizado um estudo dos sistemas de transcrição automática de laudos, no qual os principais requisitos são caracterizados e classificados. Os dois estudos acima citados foram integrados para a elaboração de uma metodologia para a avaliação de Sistemas de Transcrição Automática de Laudos em Radiologia. A metodologia foi, então, validada previamente através de inspeções e testes de usabilidade realizados fora do ambiente hospitalar, através do uso de um sistema de transcrição automática de laudos em Radiologia. Posteriormente, a metodologia foi também aplicada a um hospital da cidade de São Paulo. Como resultado principal foi proposto um guia bastante detalhado para se avaliar os Sistemas de Transcrição Automática de Laudos em Radiologia, cada vez mais presentes em hospitais e clínicas no país, além dos relatos das experiências obtidas com a aplicação desta metodologia em um caso real. / This work combines knowledge from Computer Science and Health Science in order to propose an evaluation methodology for Automatic Transcription System of Radiology Reports. At first a study regarding a Voice User Interface is presented, this interface identifies the requirements for Spoken Language Dialogue Systems and it can also be used as a tool for an evaluation methodology. Following, a study of automatic transcription systems is presented; in this study the main requirements are listed and classified. Both studies were integrated to allow a new methodology to evaluate an Automatic Transcription System in Radiology. This methodology was previously validated through some inspections and usability tests outside the hospital environment and afterword the methodology was used in a hospital in São Paulo city. As a main result it was proposed a very detailed guide for evaluating the Automatic Transcription Systems Reports in Radiology, increasingly found in hospitals and clinics in this country, apart from reports on experiences gained in applying this methodology in a real case.

[pt] A extração de parâmetros musicais de gravações de áudio a partir do processamento do sinal musical viabiliza uma série de aplicações importantes no campo de análise e classificação de peças musicais. A Harmonia é um importante aspecto da estrutura musical, fazendo da seqüência de acordes de uma peça uma informação extremamente relevante em sua análise e parte fundamental do próprio registro gráfico da música. Este trabalho discute os elementos envolvidos na transcrição de harmonia musical assim como as ferramentas matemáticas e de processamento de sinais adequadas a um método automático de análise e identificação de acordes, propondo um modelo para um sistema capaz de transcrever seqüências de acordes a partir de gravações de áudio comerciais. O sistema proposto analisa um sinal de áudio e retorna as cifras correspondentes aos acordes que melhor representam o sinal / [en] The extraction of musical parameters from audio recordings enables a series of important applications in the field of musical analysis and classification. Harmony is a major issue on music structure, making the chord sequence of a musical piece an extremely relevant information on its analysis and a fundamental part of the musical graphic register. This work discusses the elements involved in musical harmony transcription as well as the mathematical and signal processing tools suitable to build an automatic method for the analysis and identification of chords, proposing a model capable of transcribing chord sequences from commercial audio recordings. The proposed system analyzes an audio signal and returns the symbols corresponding to the chords that best fit the signal.

This thesis discusses theoretical analysis of the origin of speech, introduces applications of speech technologies and explains the contemporary approach to phonetical transcription of speech recordings. Furthermore, it describes the metrics of audio recordings quality assessment, which is split into two discrete classes. The first one groups signal quality metrics, while the other one groups content richness metrics. The first goal of the practical section is to create a statistical model for accuracy prediction of machine transcription of speech recordings based on a measurement of their quality. The second goal is to evaluate which partial metrics are the most essential for accuracy prediction of machine transcription.