Natural Language Processing (PLN) is an interdisciplinary field of research into Computing, Language and Traveller Recognition Sciences, including the use of human natural language in human-machine interaction. Most of the PLN’s investigative tasks can be implemented to solve real world problems. This is the case for the recognition and translation of natural language, which can be used to construct automated systems for the transliteration and translation of documents. For digitised hand-written documents, the transcription is used to facilitate digital access to content, as the simple digitisation of images only provides, in most cases, search by image and not by language content. Transcription is even more important for historical manuscripts, as most of these documents are unique and preservation of their content is crucial for cultural and historical reasons. The transcription of historical manuscripts is usually done by paletographs, who are former writers and vocabulary experts. Recently, the Written Recognition Systems (RES) have become a common tool to assist palletographs in their task, which provides a draft of the transcription that can be corrected by paletographs with more or less sophisticated methods. This draft transcription is useful when it presents a sufficiently low error rate to make the correction process more comfortable than a complete transcription from scratch. Therefore, obtaining a draft transcription with a low error rate is crucial for this PLN technology to be incorporated into the transcription process. The work described in this thesis focuses on improving the draft transcription offered by an RES system, with the aim of reducing the effort made by the paletographs to obtain the transcription of digitised historical manuscripts. This problem is faced on the basis of three different but complementary scenarios: · Multimodality: The use of RES allows the paletographs to accelerate the manual transcription process as they are able to correct in a draft transcription. Another alternative is to obtain the draft transcript by dictating the content to a Habla Automatic Recognition System. When both sources are available, a multimodal combination of these sources is possible and an iterative process can be carried out to refine the final scenario. · Interactivity: The use of assistive technologies in the transcription process makes it possible to reduce the time and human effort required to obtain the correct transcription, thanks to the cooperation between the care system and the palaetograph to obtain perfect transcription. Multi-modal feedback can be used in the care system to provide additional sources of information with signals representing the same sequence of words to be transcribed (e.g. text image, or speech signal of the content of that text image), or signals representing only one word or character to be corrected (e.g. handwritten word using a touch screen). · Crowdsourcing: Distributed and open collaboration emerge as a powerful tool for mass transcription at a relatively low cost, as the monitoring effort of the paletographs can be drastically reduced. The multimodal combination makes it possible to use the dictation of handwritten text line content on a multi-modal crowdsourcing platform, where contributors can provide speech samples using their own mobile device instead of using computers.