Colloquial Finnish in Subtitles – a Survey on the British Situation Comedy Peep Show

The use of dialects in audiovisual texts is increasing, but the use of standard language in Finnish subtitles is still the prevalent norm and seldom flouted. This may not be in the best interest of the audience as the way the characters speak – their idiolects – is very important for the atmosphere of the audiovisual text and widely used to give the characters personalities and background. This thesis studies whether the viewing experience could be enhanced by taking the characters’ idiolects into account in subtitling. The study was executed as a survey, and the respondents were recruited from social media and an university of applied sciences. A total of 113 respondents were divided into two groups: the experimental group (n=59) and the control group (n=54). Both groups were shown an excerpt of the British situation comedy Peep Show, but with different subtitles. In the experimental subtitles, three characters were given a written idiolect with a level of colloquial language corresponding to the spoken idiolect of the character, while the control subtitles followed the norm of using standard language. The questionnaire contained background questions, a Likert-scale question and open questions. The quantitative responses were analysed statistically through cross tabulation and Mann–Whitney U test or Kruskal-Wallis test (CI=95%, α=0,05). The results showed a statistically significant difference in keeping track on which of the characters was speaking for the benefit of the experimental subtitles. In the other items no statistically significant difference was found between the groups. In the open questions the use of colloquial language was mostly commented favourably.

Sentence Compression with Supervised Machine Learning

The subject of the thesis is automatic sentence compression with machine learning, so that the compressed sentences remain both grammatical and retain their essential meaning. There are multiple possible uses for the compression of natural language sentences. In this thesis the focus is generation of television program subtitles, which often are compressed version of the original script of the program. The main part of the thesis consists of machine learning experiments for automatic sentence compression using different approaches to the problem. The machine learning methods used for this work are linear-chain conditional random fields and support vector machines. Also we take a look which automatic text analysis methods provide useful features for the task. The data used for machine learning is supplied by Lingsoft Inc. and consists of subtitles in both compressed an uncompressed form. The models are compared to a baseline system and comparisons are made both automatically and also using human evaluation, because of the potentially subjective nature of the output. The best result is achieved using a CRF - sequence classification using a rich feature set. All text analysis methods help classification and most useful method is morphological analysis. Tutkielman aihe on suomenkielisten lauseiden automaattinen tiivistäminen koneellisesti, niin että lyhennetyt lauseet säilyttävät olennaisen informaationsa ja pysyvät kieliopillisina. Luonnollisen kielen lauseiden tiivistämiselle on monta käyttötarkoitusta, mutta tässä tutkielmassa aihetta lähestytään television ohjelmien tekstittämisen kautta, johon käytännössä kuuluu alkuperäisen tekstin lyhentäminen televisioruudulle paremmin sopivaksi. Tutkielmassa kokeillaan erilaisia koneoppimismenetelmiä tekstin automaatiseen lyhentämiseen ja tarkastellaan miten hyvin erilaiset luonnollisen kielen analyysimenetelmät tuottavat informaatiota, joka auttaa näitä menetelmiä lyhentämään lauseita. Lisäksi tarkastellaan minkälainen lähestymistapa tuottaa parhaan lopputuloksen. Käytetyt koneoppimismenetelmät ovat tukivektorikone ja lineaarisen sekvenssin mallinen CRF. Koneoppimisen tukena käytetään tekstityksiä niiden eri käsittelyvaiheissa, jotka on saatu Lingsoft OY:ltä. Luotuja malleja vertaillaan Lopulta mallien lopputuloksia evaluoidaan automaattisesti ja koska teksti lopputuksena on jossain määrin subjektiivinen myös ihmisarviointiin perustuen. Vertailukohtana toimii kirjallisuudesta poimittu menetelmä. Tutkielman tuloksena paras lopputulos saadaan aikaan käyttäen CRF sekvenssi-luokittelijaa laajalla piirrejoukolla. Kaikki kokeillut teksin analyysimenetelmät auttavat luokittelussa, joista tärkeimmän panoksen antaa morfologinen analyysi.

Marvel Cinematic Universe in translation: a study of intertextual coherence and inter-filmic discourse

The primary objective of the present thesis was to determine the extent of intertextual coherence and inter-filmic discourse retained in the Finnish DVD subtitles of the first twelve feature films set in the Marvel Cinematic Universe and ten episodes of the first season of Agents of S.H.I.E.L.D., a transmedia extension of the MCU. The cinematic world of Marvel was chosen as research data in this study for the inherence, abundance, and conspicuousness of its intertextuality. Two categories through which to retain intertextual coherence in translation were set as the premise of the study: 1) the consistent application of the same form of MCU-related proper names in translation and 2) the retention of MCU-related allusions in translation when the retention of the allusion is a strategic choice. The data was collected and analyzed primarily in this juxtaposition. The examination of the gathered data and the set research questions necessitated the division of audiovisual allusions into three categories: verbal visual allusions, secondary spoken allusions, and primary spoken allusions, the last of which was further divided into ambiguous and unambiguous types. Because of their qualitative inadequacies, unambiguous primary spoken allusions were not eligible as data in the present study. 33.3 percent of the proper names qualified as data were translated consistently in each installment they were referenced. In terms of allusions, 76.2 percent of the qualified source-text instances were retained in translation. The results indicate that intertextual elements are more easily identified and retained within the context of one narrative than when this requires the observation of multiple connected narratives as one interwoven universe.

“I haven’t had that much fun since labour.” Subtitling sarco-ironic remarks in Gilmore Girls

The purpose of this study is to explore how the potential cues of spoken sarcasm and irony are transferred into written Finnish subtitles in the American television series Gilmore Girls. The aim is to discover how the use of cuing differs between the English source text and the Finnish target text. The research is conducted through qualitative and quantitative analysis and comparison between the two languages. Three (3) episodes from Gilmore Girls are analysed using the Finnish DVD subtitles and the episodes’ English unofficial transcripts. The sarco-ironic remarks and their translations are identified, and their individual sarco-ironic cues are categorised. The number of cues, the number of cues in each category and the number of cue shifts are compared quantitatively in order to find whether the translator uses less or different cues than the writer of the original source text. The qualitative analysis focuses on the cue types used in both texts. The results confirmed the hypothesis of sarcasm and irony being relatively straightforward linguistic phenomena to translate, and the hypothesis of the category of words of reinforcement/trivialising being the most used in conveying sarcasm and irony. However, the study also showed that even though some cues were cut in the translation process, the sarco-ironic meaning was rarely omitted entirely from the translation. The findings also showed that the number of cue shifts was surprisingly low. From these results we can deduce that at least in the context of DVD subtitles, sarcasm and irony can be conveyed also in writing, and even though some cues were omitted or shifted, sarcasm and irony were still detectable in the Finnish subtitles. The language used in DVD translations uses less cues of sarcasm and irony and it could thus be considered less versatile than the original source language.