Wong Chi Hang. / Thesis submitted in: October 2005. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2006. / Includes bibliographical references (leaves 89-94). / Abstracts in English and Chinese. / Abstract --- p.ii / 摘要 --- p.iii / Acknowledgement --- p.iv / Chapter 1 --- Introduction --- p.1 / Chapter 2 --- Literature Review --- p.5 / Chapter 2.1 --- LyricAlly --- p.5 / Chapter 2.2 --- Singing Voice Detection --- p.6 / Chapter 2.3 --- Singing Transcription System --- p.7 / Chapter 3 --- Background and System Overview --- p.9 / Chapter 3.1 --- Background --- p.9 / Chapter 3.1.1 --- Audio Mixing Practices of the popular music industry --- p.10 / Chapter 3.1.2 --- Cantonese lyric writer practice --- p.11 / Chapter 3.2 --- System Overview --- p.13 / Chapter 4 --- Vocal Signal Enhancement --- p.15 / Chapter 4.1 --- Method --- p.15 / Chapter 4.1.1 --- Non-center Signal Estimation --- p.16 / Chapter 4.1.2 --- Center Signal Estimation --- p.17 / Chapter 4.1.3 --- Bass and drum reduction --- p.21 / Chapter 4.2 --- Experimental Results --- p.21 / Chapter 4.2.1 --- Experimental Setup --- p.21 / Chapter 4.2.2 --- Results and Discussion --- p.24 / Chapter 5 --- Onset Detection --- p.29 / Chapter 5.1 --- Method --- p.29 / Chapter 5.1.1 --- Envelope Extraction --- p.30 / Chapter 5.1.2 --- Relative Difference Function --- p.32 / Chapter 5.1.3 --- Post-Processing --- p.32 / Chapter 5.2 --- Experimental Results --- p.34 / Chapter 5.2.1 --- Experimental Setup --- p.34 / Chapter 5.2.2 --- Results and Discussion --- p.35 / Chapter 6 --- Non-vocal Pruning --- p.39 / Chapter 6.1 --- Method --- p.39 / Chapter 6.1.1 --- Vocal Feature Selection --- p.39 / Chapter 6.1.2 --- Feed-forward neural network --- p.44 / Chapter 6.2 --- Experimental Results --- p.46 / Chapter 6.2.1 --- Experimental Setup --- p.46 / Chapter 6.2.2 --- Results and Discussion --- p.48 / Chapter 7 --- Lyric Feature Extraction --- p.51 / Chapter 7.1 --- Features --- p.52 / Chapter 7.1.1 --- Relative Pitch Feature --- p.52 / Chapter 7.1.2 --- Time Distance Feature --- p.54 / Chapter 7.2 --- Pitch Extraction --- p.56 / Chapter 7.2.1 --- f0 Detection Algorithms --- p.56 / Chapter 7.2.2 --- Post-Processing --- p.64 / Chapter 7.2.3 --- Experimental Results --- p.64 / Chapter 8 --- Lyrics Alignment --- p.69 / Chapter 8.1 --- Dynamic Time Warping --- p.69 / Chapter 8.2 --- Experimental Results --- p.72 / Chapter 8.2.1 --- Experimental Setup --- p.72 / Chapter 8.2.2 --- Results and Discussion --- p.74 / Chapter 9 --- Conclusion and Future Work --- p.82 / Chapter 9.1 --- Conclusion --- p.82 / Chapter 9.2 --- Future Work --- p.83 / Chapter A --- Publications --- p.85 / Chapter B --- Symbol Table --- p.86 / Bibliography --- p.89

M.A. (Applied Linguistics and Literary Theory) / Fan subtitling, or fansubbing, has become a new form of subtitling, in which fans, often with little to no training, subtitle a film or series they enjoy, and then release the subtitles or a subtitled version of the film on the internet for free download, often by other fans, who often have an interest in the source language and culture and who may have an interest in using subtitles as a language-learning device. Subtitles produced for this fan audience can differ substantially from those created in a professional environment. For this reason, fansubs could be seen as a new genre of subtitling, employing strategies that would be unacceptable in a professional subtitled product aimed at general consumption. Professional subtitlers and translators are encouraged to follow the norms or ‘best practice’ of their industry in order to produce subtitles that are consistent, of a high quality and that have been made easy to process through methods such as length reduction and domestication. Fansubbers have no such imperatives imposed on their subtitles, being mostly self-taught and often judged by their peers and viewers on the ‘authenticity’ of their translations, rather than by other professionals and their standards. This creates an environment in which more creative translation strategies and presentation become possible, in order to create a translation that will be approved of by fans with a special interest in linguistic and cultural content. This need for the appearance of faithfulness or authenticity seems to result in a more source-oriented and complete translation approach, and a corpus-based investigation of French fansubs appears to confirm this. This dissertation attempts to explore the differences found in the two genres by analysing and comparing a translational corpus of professionally-produced and fanproduced English subtitles of French films. The average length of the subtitles and instances of discourse markers in the two sets of subtitles are compared in order to determine the degree of reduction found in each type of subtitle, and from the source to target culture. The translation strategies used in the two types of subtitling differ, with the fansub opting for a more complete rendering of the source text dialogue and a more source-oriented approach, which retains elements of spoken language found in the film dialogue. It appears that the differences between the target texts may be influenced by very different contexts in which they are produced, and possibly the differing needs and expectations of the audiences of the two genres.

University of Macau / Faculty of Social Sciences and Humanities / Department of English