Upload
lillian-mcmahon
View
218
Download
1
Tags:
Embed Size (px)
Citation preview
UCCTS 2010, 27.-29.7.2010
The dilemma between corpus statistics and reception of a text: An analysis of foreignising and domesticating elements of translations
Hannu Kemppanen, Jukka Mäkisalo & Grigory GurinUniversity of Eastern Finland
Venuti (1995) criticism (Tymoczko 2000, Boyden 2006)- obscurity of the notions- dichotomy Attempts to concretise the concepts- e.g. Pedersen 2005
keyword studies- comparing translated and non-translated texts- keywords as untypical, foreign elements (Kemppanen 2004,
2008) study where statistical features of translated texts were
compared with the results of an evaluation test (Kemppanen and Mäkisalo 2010)
- no correlation between the statistical features and the results of the test
- subjectivity in ranking translations- individual words/phrases and foreign elements draw
subjects’ attention
possible correlation between statistical features of the texts and the results of the evaluation test a corpus-based analysis of non-fiction translations (Russian-Finnish) and non-translations- foreignising/domesticating features of translated vs. non-translated texts an evaluation test- foreignising/domesticating features of translated vs. non-translated texts (cf. the former study: different subjects, different reference
corpus)
• foreignising/domesticating features of translated vs. non-translated texts
foreignising/domesticating features of translated vs. non-translated texts
*
Keywords:- number of keywords- keyness maximum value- keyness mean value Other features:- type/token ratio- mean length of sentences
k
RESULTS: STATISTICAL FEATURES There are only weak statistical
correlations between some of the features.
On one hand, type/token ratio correlates to some extent reversely with mean keyness value (Pearson p = 0,62).
On the other hand, weak correlation between a high number of keywords and high keyness maximum value (p = 0,51).
However, overall, when history texts are compared to newspaper texts, various statistical features do not correlate with each other.
BartenjevKomissarov
HolodkovskijBaryshnikov
ApunenTarkka
0
50
100
150
200
250
300
350
400
NumberKeyws
KeynMax
KeynMean
TTR
SentLength
A questionnaire for ranking the texts according to (subjective) impression of domestication/ foreignisation.
Evaluating extracts (1000 words) of four Russian–Finnish translations and two non-translations on Finnish political history on a scale 1–5 (domestic–foreign).
In addition, naming at least one foreignising or domesticating feature in each text
Pilot: five subjects, translation students (earlier six translation trainers)
Four Russian-Finnish translations and two non-translated Finnish history texts were ranked according to the median of evaluations
The ranges of evaluations between the texts varied a lot, highlighting the difference between translations and non-translations.
EvalMed Range Tr/Non-tr
Tarkka 2 1 – 3 2/3 Apunen 3 1 – 3 1/4 Bartenjev 3 2 – 5 3/2 Holodkovskij 4 3 – 4 4/1 Komissarov 4 4 – 5 5/0 Baryshnikov 4 4 – 5 5/0
Results: statistical features and the evaluation test
The ranking of evaluation correlates only weakly with sentence length (p = 0,59).
With various keyness values or TTR, the evaluation test has no correlation.
BartenjevKomissarov
HolodkovskijBaryshnikov
ApunenTarkka
0
50
100
150
200
250
300
350
400
NumberKeyws
KeynMax
KeynMean
TTR
SentLength
EvalMed
Sentence structure foreign (10) Word order foreign (2) Phrases/Collocations foreign (4) Phrases/Collocations domestic (colourful
expressions) (3) Individual words foreign (4) (adjectives)
Attitudinal features foreign (10) (NB: foreign point of view in a fluent text, one comment)
Attitudinal features domestic (2) (point of view, neutrality)
Fluency/style domestic (7) (fluent/good Finnish) Non-fluency/style foreign (1) Orthography foreign (1) Explanations foreign (translation) (1)
results of the study support the earlier empirical findings
- for the most part, statistical features of the texts do not correlate with the results of the evaluation test
- various statistical features retrieved from the corpus analysis are not in line with each other
new findings- the evaluation test differentiates non-translated texts
from translated texts, and furthermore, more detailed sub-groups of translations
Can a translation be recognised on the grounds of the analysed features?
- on the grounds of statistical features – NO- on the grounds of the evaluation test – YES Categorisation of texts into translated and non-
translated texts, and naming of text features in a qualitative study suggest that foreigness of a text is a marked feature
Thank you!Questions?Comments?