O som no cinema gera expectativas, guia o espectador através das imagens e molda a recepção das cenas (BORDWELL; THOMPSON, 2008). Para que o espectador surdo ou ensurdecido tenha acesso a esse input ao assistir a uma produção audiovisual, a legenda para surdos e ensurdecidos (LSE) traduz os efeitos sonoros dos filmes (ARAÚJO, 2004). Em busca de um modelo de LSE que supra as necessidades do público alvo, foi realizada uma pesquisa com 34 surdos brasileiros (ARAÚJO; NASCIMENTO, 2011; ARAÚJO ET AL, 2013). Os resultados sugeriram que as traduções existentes analisadas de efeitos sonoros confundiam o espectador ou não acrescentavam informações relevantes. Levando em consideração a importância do som no cinema, realizou-se uma pesquisa para observar como essas traduções ocorriam em três filmes brasileiros comercializados em DVD (NASCIMENTO, 2013). Os resultados mostraram que a maioria dos efeitos sonoros encontrados nos filmes da pesquisa foram legendados sem uma preocupação aparente de ligar o som à sua significação na trama. Assim, com o propósito de auxiliar os legendistas nessas traduções, esta pesquisa tem como objetivo principal convencionar legendas para efeitos sonoros, além de estabelecer algumas diretrizes para a tarefa de traduzir sons em palavras. Acredita-se, ainda, que o processo de padronização das legendas também tornaria mais acessível as legendas em língua portuguesa para o público-alvo, pois unidades convencionadas são lidas de forma mais rápida (WOOD, 2015). Para isso, foi coletado um corpus comparável que foi analisado com o auxílio do software WordSmith Tools. Para facilitar as análises com o software, as legendas dos filmes foram etiquetadas a partir de categorias de proveniência do som. Para os ruídos, apresentamse as seguintes etiquetas: sons produzidos pelo homem, sons produzidos por objeto, sons produzidos por animais, sons produzidos pela natureza, sons ficcionais e silêncio. Para as músicas, as etiquetas são: música de fosso e música de tela. Essas etiquetas foram utilizadas como nódulos de busca e possibilitaram encontrar relevantes dados numéricos, tais como as categorias mais frequentemente legendadas no corpus: sons produzidos pelo homem e música de fosso. Portanto, esses são os sons cujas traduções foram convencionadas nessa pesquisa. / The sound in the cinema creates expectations, guides the spectator through images, and shapes the reception of scenes (BORDWELL; THOMPSON, 2008). So that deaf and hard of hearing public may have access to this input when watching an audiovisual production, the subtitle for the deaf and hard of hearing (SDH) translates the movies sound effects (ARAÚJO, 2004). In search for a SDH model that meets the spectators needs, a research has been conducted with 34 Brazilians deaf subjects (ARAÚJO; NASCIMENTO, 2011; ARAÚJO ET AL, 2013). The results suggested that the existent sound effects translations would mislead the spectator or would not contribute with relevant information. Taking into account the importance of the sound in the cinema, another research has been conduct in order to analyze how translations of sound effect were made in three commercialized DVDs in Brazil (NASCIMENTO, 2013). The results showed that most sound effects in the films were translated without connecting them to their meanings within the movie plot. Thus, in order to help translators with these translations, this researchs main objective is to conventionalize sound effects subtitles and stablish some guidelines to the sound into words task. Furthermore, it is believed that conventionalizing subtitles could render them more accessible as conventionalized units are read faster than non-conventionalized (WOOD, 2015). Therefore, a comparable corpus was collected and analyzed by means of WordSmith Tools software. In order to render the analysis simpler, the corpus was tagged following categories defined by the sound provenience. The following tags were applied to noise: sound produced by men, sound produced by objects, sound produced by nature, sound produced by animals, fictional sounds and silence. To music, the tags were background music and screen music. These tags were used as search words and rendered possible find relevant data, such as the most frequently translated categories within the corpus, which are: sounds produced by men and background music. Therefore, these are the sound categories whose translations were conventionalized in this research.

O som no cinema gera expectativas, guia o espectador através das imagens e molda a recepção das cenas (BORDWELL; THOMPSON, 2008). Para que o espectador surdo ou ensurdecido tenha acesso a esse input ao assistir a uma produção audiovisual, a legenda para surdos e ensurdecidos (LSE) traduz os efeitos sonoros dos filmes (ARAÚJO, 2004). Em busca de um modelo de LSE que supra as necessidades do público alvo, foi realizada uma pesquisa com 34 surdos brasileiros (ARAÚJO; NASCIMENTO, 2011; ARAÚJO ET AL, 2013). Os resultados sugeriram que as traduções existentes analisadas de efeitos sonoros confundiam o espectador ou não acrescentavam informações relevantes. Levando em consideração a importância do som no cinema, realizou-se uma pesquisa para observar como essas traduções ocorriam em três filmes brasileiros comercializados em DVD (NASCIMENTO, 2013). Os resultados mostraram que a maioria dos efeitos sonoros encontrados nos filmes da pesquisa foram legendados sem uma preocupação aparente de ligar o som à sua significação na trama. Assim, com o propósito de auxiliar os legendistas nessas traduções, esta pesquisa tem como objetivo principal convencionar legendas para efeitos sonoros, além de estabelecer algumas diretrizes para a tarefa de traduzir sons em palavras. Acredita-se, ainda, que o processo de padronização das legendas também tornaria mais acessível as legendas em língua portuguesa para o público-alvo, pois unidades convencionadas são lidas de forma mais rápida (WOOD, 2015). Para isso, foi coletado um corpus comparável que foi analisado com o auxílio do software WordSmith Tools. Para facilitar as análises com o software, as legendas dos filmes foram etiquetadas a partir de categorias de proveniência do som. Para os ruídos, apresentamse as seguintes etiquetas: sons produzidos pelo homem, sons produzidos por objeto, sons produzidos por animais, sons produzidos pela natureza, sons ficcionais e silêncio. Para as músicas, as etiquetas são: música de fosso e música de tela. Essas etiquetas foram utilizadas como nódulos de busca e possibilitaram encontrar relevantes dados numéricos, tais como as categorias mais frequentemente legendadas no corpus: sons produzidos pelo homem e música de fosso. Portanto, esses são os sons cujas traduções foram convencionadas nessa pesquisa. / The sound in the cinema creates expectations, guides the spectator through images, and shapes the reception of scenes (BORDWELL; THOMPSON, 2008). So that deaf and hard of hearing public may have access to this input when watching an audiovisual production, the subtitle for the deaf and hard of hearing (SDH) translates the movies sound effects (ARAÚJO, 2004). In search for a SDH model that meets the spectators needs, a research has been conducted with 34 Brazilians deaf subjects (ARAÚJO; NASCIMENTO, 2011; ARAÚJO ET AL, 2013). The results suggested that the existent sound effects translations would mislead the spectator or would not contribute with relevant information. Taking into account the importance of the sound in the cinema, another research has been conduct in order to analyze how translations of sound effect were made in three commercialized DVDs in Brazil (NASCIMENTO, 2013). The results showed that most sound effects in the films were translated without connecting them to their meanings within the movie plot. Thus, in order to help translators with these translations, this researchs main objective is to conventionalize sound effects subtitles and stablish some guidelines to the sound into words task. Furthermore, it is believed that conventionalizing subtitles could render them more accessible as conventionalized units are read faster than non-conventionalized (WOOD, 2015). Therefore, a comparable corpus was collected and analyzed by means of WordSmith Tools software. In order to render the analysis simpler, the corpus was tagged following categories defined by the sound provenience. The following tags were applied to noise: sound produced by men, sound produced by objects, sound produced by nature, sound produced by animals, fictional sounds and silence. To music, the tags were background music and screen music. These tags were used as search words and rendered possible find relevant data, such as the most frequently translated categories within the corpus, which are: sounds produced by men and background music. Therefore, these are the sound categories whose translations were conventionalized in this research.

<p>The last ten years have been the witnesses of the emergence of any kind of video content. Moreover, the appearance of dedicated websites for this phenomenon has increased the importance the public gives to it. In the same time, certain individuals are deaf and occasionally cannot understand the meanings of such videos because there is not any text transcription available. Therefore, it is necessary to find solutions for the purpose of making these media artefacts accessible for most people. Several software propose utilities to create subtitles for videos but all require an extensive participation of the user. Thence, a more automated concept is envisaged. This thesis report indicates a way to generate subtitles following standards by using speech recognition. Three parts are distinguished. The first one consists in separating audio from video and converting the audio in suitable format if necessary. The second phase proceeds to the recognition of speech contained in the audio. The ultimate stage generates a subtitle file from the recognition results of the previous step. Directions of implementation have been proposed for the three distinct modules. The experiment results have not done enough satisfaction and adjustments have to be realized for further work. Decoding parallelization, use of well trained models, and punctuation insertion are some of the improvements to be done.</p>

This dissertation is a comparative investigation of the work of two subtitlers of Swedish films translated into French in the last decade of 20th century. It is a look at how they try to render the illusion of spoken language in the written line that is a subtitle. This effect has to be created despite the constraints of time, synchronisation, readability, and the limited number of characters available to work with. We have shown that the subtitles examined in this work include many of the characteristics typical of spoken French. These characteristics can be classified as belonging to the Phonetic, the Lexical or the Syntactic level. At the lexical level we can see that the translators use expressions with a lower style-level than the originals but manage to avoid being too shocking, innovative or destabilizing for the audience. We have also shown diastratic language differences in translation: our subtitlers strengthen characters of social differences by using a different language style depending on the status that the person in question has. Because of the polysemiotic nature of film it is not the primary purpose of a subtitler to translate everything. Typically verbal speech characteristics such as repetition, hesitation, incomplete phrases and non-informational expressions that are often seen as less acceptable in the written word are normally sacrificed in subtitles. On the other hand, we could not show that the original language (Swedish) had a sizable impact on the style and construction of the subtitles. We also looked at possible stylistic differences between the two subtitlers translations and tried to find out if the “voice” of the subtitler shone through. If so, we were interested in attempting to see if a difference could be caused by the fact that one of the subtitlers has French as her native language while the other translates from his second language. We were not able to relate the cause of the difference of style we observed to anything other than the original dialogue’s changing style-levels. In which direction the verbal content in subtitles will develop is an interesting and relevant question. In this work we have identified a development of style that has taken place during recent years in French subtitles to Swedish films. This development goes from a previous situation where the style of the spoken word was practically unused compared to today’s more frequent and systematic use.

The last ten years have been the witnesses of the emergence of any kind of video content. Moreover, the appearance of dedicated websites for this phenomenon has increased the importance the public gives to it. In the same time, certain individuals are deaf and occasionally cannot understand the meanings of such videos because there is not any text transcription available. Therefore, it is necessary to find solutions for the purpose of making these media artefacts accessible for most people. Several software propose utilities to create subtitles for videos but all require an extensive participation of the user. Thence, a more automated concept is envisaged. This thesis report indicates a way to generate subtitles following standards by using speech recognition. Three parts are distinguished. The first one consists in separating audio from video and converting the audio in suitable format if necessary. The second phase proceeds to the recognition of speech contained in the audio. The ultimate stage generates a subtitle file from the recognition results of the previous step. Directions of implementation have been proposed for the three distinct modules. The experiment results have not done enough satisfaction and adjustments have to be realized for further work. Decoding parallelization, use of well trained models, and punctuation insertion are some of the improvements to be done.

When subtitles are burned into a video, an error can sometimes occur in the encoder that results in the same subtitle being burned into several frames, resulting in subtitles becoming frozen. This thesis provides a way to detect frozen video subtitles with the help of an implemented text detector and classifier. Two types of classifiers, naïve classifiers and machine learning classifiers, are tested and compared on a variety of different videos to see how much a machine learning approach can improve the performance. The naïve classifiers are evaluated using ground truth data to gain an understanding of the importance of good text detection. To understand the difficulty of the problem, two different machine learning classifiers are tested, logistic regression and random forests. The result shows that machine learning improves the performance over using naïve classifiers by improving the specificity from approximately 87.3% to 95.8% and improving the accuracy from 93.3% to 95.5%. Random forests achieve the best overall performance, but the difference compared to when using logistic regression is small enough that more computationally complex machine learning classifiers are not necessary. Using the ground truth shows that the weaker naïve classifiers would be improved by at least 4.2% accuracy, thus a better text detector is warranted. This thesis shows that machine learning is a viable option for detecting frozen video subtitles.

O objetivo deste trabalho é analisar como é feita a legendagem utilizada pelos canais de televisão por assinatura no Brasil e verificar se é possível identificar um padrão comum a todos eles. A utilização da proposta de padrão de legendagem para a Europa apresentada por Fotios Karamitroglou e a minha experiência na área como tradutora de legendas para TV serviram como base para realizar as análises. O estudo utiliza corpora em língua inglesa, as falas originais dos filmes, e em língua portuguesa, a transcrição das legendas de filmes exibidos pelos canais. A escolha dos canais considerou a relevância que a legendagem tem em sua programação diária. A análise foi feita por meio de comparações entre o corpus da transcrição do original e o das legendas e foram considerados tanto os aspectos técnicos da legendagem quanto os textuais e tradutórios. Os resultados foram compilados e podem servir de base para analisar outros aspectos da tradução para legendas. / This research aims to analyze the Brazilian Cable Channels Subtitling in order to verify if it is possible to identify a standard for subtitling. \"A Proposed Set for Subtitling Standard in Europe\" by Fotios Karamitroglou, and my experience as a translator are the basis for doing the analysis. The study uses an English corpus, the original lines from the movies, and a Portuguese corpus, the subtitle transcription from the movies broadcasted by the channels. The channels selection took account of the relevance that subtitling has in their daily schedule. The analysis was done by comparing both corpora, and technical, textual and translation aspects, and the final results was compiled and can be used as a base for analyze another subtitling aspects.

O objetivo deste trabalho é analisar como é feita a legendagem utilizada pelos canais de televisão por assinatura no Brasil e verificar se é possível identificar um padrão comum a todos eles. A utilização da proposta de padrão de legendagem para a Europa apresentada por Fotios Karamitroglou e a minha experiência na área como tradutora de legendas para TV serviram como base para realizar as análises. O estudo utiliza corpora em língua inglesa, as falas originais dos filmes, e em língua portuguesa, a transcrição das legendas de filmes exibidos pelos canais. A escolha dos canais considerou a relevância que a legendagem tem em sua programação diária. A análise foi feita por meio de comparações entre o corpus da transcrição do original e o das legendas e foram considerados tanto os aspectos técnicos da legendagem quanto os textuais e tradutórios. Os resultados foram compilados e podem servir de base para analisar outros aspectos da tradução para legendas. / This research aims to analyze the Brazilian Cable Channels Subtitling in order to verify if it is possible to identify a standard for subtitling. \"A Proposed Set for Subtitling Standard in Europe\" by Fotios Karamitroglou, and my experience as a translator are the basis for doing the analysis. The study uses an English corpus, the original lines from the movies, and a Portuguese corpus, the subtitle transcription from the movies broadcasted by the channels. The channels selection took account of the relevance that subtitling has in their daily schedule. The analysis was done by comparing both corpora, and technical, textual and translation aspects, and the final results was compiled and can be used as a base for analyze another subtitling aspects.

網路社群中的字幕組，是由一群無償勞動的網友們、透過新科技溝通媒介而組織起來的。在本研究中，筆者將字幕組視為一個虛擬團隊，意圖探討無償虛擬團隊的組織文化與互動方式。近年來，國內的虛擬團隊研究觀點，似乎較偏向學校或者企業，所設定的虛擬團隊也清一色是營利組織或者學校體制內的虛擬團隊。然而，字幕組的工作卻是無償無薪，也缺乏外在的企業體制。 無償的虛擬團隊與營利組織裡的虛擬團隊有何差異呢？網路社群中的P2P資源交換特性或者網路社群中的生態，是否也會影響到字幕組這樣的無償虛擬團隊？本研究採取Potter與Balthazard（2002）所整理出之團隊「互動型態」理論、Dani 與Burns等人（2006）的「組織文化類型」理論，試圖找出不同規模與文化情境的字幕組互動型態、組織文化類型。筆者觀察了數個中國與台灣的字幕組，期望找出具集體知識共作、創造特性的虛擬團隊特色。而這些特色，或許能啟發國內其餘具知識分享、文化創造性質的虛擬團隊。 本研究使用參與觀察、親身實作等方法進行，輔以筆者自身的中國字幕組經驗與其他台灣字幕組成員的深度訪談，同步進行書寫。 研究結果顯示，字幕組的組織文化偏向關係親密的「宗族文化」、以及具有創新特質的「鬆散彈性文化」。而針對字幕組的互動型態分析，則集中於「合作型」、「人際型」，同樣都是關係較親近、競爭壓力較少的互動類型。這些特質顯示出，字幕組這樣的無償虛擬團隊，的確在組織文化、互動風格方面，都有著與企業不同的組織屬性。 / The Subtitle Teams in online communities are formed by volunteers who communicate by virtual technology. This study views the Subtitle Teams as virtual teams, in order to explore their organizational culture and interaction style. In recent years, most Virtual Teams studies in Taiwan focus on teams in school and businesses industries, however, the Subtitle Teams are volunteers without business structure. What is the difference between commercial virtual teams and voluntary subtitle teams? Do P2P communities influence voluntary subtitle teams like Subtitle Teams? We use Potter & Balthazard (2002)’s Interaction Style theory and Dani & Burn(2006)’s Organization Culture Types to explore the traits of Subtitle Teams. The Subtitle Teams in this study includes culture difference between Taiwan and Chinese, also the difference between Drama Subtitle Teams and Animation Subtitle Teams. The goal of this study is to find out some trait of creative, knowledge-based, and cooperative virtual teams, hope these result will help other virtual teams in NPO organizations or Cultural Industry in Taiwan. This study use participant observation, in-depth interviewing and case study to view the experience of the members from four Subtitle Teams. We found out that the Subtitle Team’s Organization Culture might tend to be The Clan Culture and The Adhocracy Culture. As to the Interaction Style analysis, many subtitle teams tend to be Passive Style and Constructive Style at the same time at the same time.

The master's thesis deals with simultaneous film interpreting and pre-prepared electronic subtitles at film festivals in the Czech Republic. Simultaneous film interpreting used to be a widespread mode of audiovisual translation at film festivals. However, today it has been mostly replaced with subtitles; these are often pre-prepared electronic subtitles manually projected under the film screen by a so-called subtitle launcher. These types of audiovisual translation are generally characterized in the theoretical part of the thesis. In the empirical part, the research first focuses on film interpreting at the Karlovy Vary International Film Festival. As films are no longer interpreted at the festival, the research method of interviews with the then interpreters was chosen. Based on the information obtained from the interviews, it is described how films were interpreted and what working conditions interpreters had. General information on film interpreting was also gathered during the interviews. The second chapter of the empirical part deals with pre-prepared subtitles: first with the process of subtitle launching and subtitle launchers' working conditions, and then with the audience's opinion, which was elicited in a questionnaire survey at the Summer Film School Uherske Hradiste. The questionnaire...