• Aucun résultat trouvé

86 Campbell et al., column Y.

0 0.5 1 1.5 2 2.5 3 3.5 4

Baseline English Translated BCV

Quantifier usage as %

55

lower overall clarity than the equivalent texts that have not passed through the filter of the translation process.

Numbers

Both translated and untranslated shareholder letters show a lower usage rate of number words than the baseline rate calculated by Pennebaker, a result that would seem surprising at first glance given the rather prominent role of numbers in annual reporting in general, and especially for companies within the financial sector. This is a category for which the LIWC results point to almost no difference at all between the English–original texts and those translated from French, with usage rates of 1.32% and 1.37%

respectively, compared to the baseline usage rate of 2.7%, twice as high. At least one simple observation shows that in this instance, Pennebaker’s dictionary would need editing in order to provide a more accurate assessment of the use of number words in shareholder letters, and indeed in financial reporting texts in general: while he has not failed to include words such as “million*” and “billion*”, his dictionary does fail to account for the common abbreviations applied to such figures in financial reporting documents.

For that reason, the commonly–observed (in financial annual reports, at least) use of constructions such as “126m” for one hundred and twenty–six million, or “1.1bn” for one point one billion would not register through LIWC as hits for this category and would therefore fail to be considered.

The particularly low usage rate–only 0.76%–recorded in the translations from BCV, which personal experience shows to use the number abbreviations mentioned above, reinforces this interpretation of these results overall. In this case, it is arguably not even necessary to conduct a concordance analysis to confirm this missing evidence, given the prevalence of these abbreviations in the kinds of texts comprising our corpora; instead, this category serves as a reminder that Pennebaker’s categories, although carefully vetted for representativeness against the language in general, may have shortcomings regarding some specific linguistic phenomena of interest to the financial translator, and so may only be a first step rather than a comprehensive language analysis tool regarding some aspects of the language used in certain textual genres.

Swear words

Pennebaker’s final linguistic process category, swear words, generates zero hits across all our shareholder letters. It is not surprising that scatological and blasphemous

56

language constitutes a taboo in our professionally produced, high–level texts, and we will accept this perfectly clear result – mirrored by a result of 0.00 at BCV as well – and move on to Pennebaker’s second overarching grouping of categories, psychological processes.

Psychological Processes

Pennebaker’s categories of words and word stems related to what he calls “psychological processes” collect broad swathes of words, both content words and function words, which he has associated with various processes through questionnaires and surveys that have been checked and rechecked over years for consistency, reliability, and statistical relevance. Nevertheless, the specific content mirrored by the words in these categories means that they are often of less direct interest to a study of financial–sector texts than were the linguistic categories which, being rooted in language rather than themes, generally had at least some application to our selection of texts. For that reason, we will examine our LIWC results for the categories in the “Psychological Processes” group from a step higher in Pennebaker’s category hierarchy, leaving any specificities of note from his sub–categories for comment in the context of the main categories: Social, Affective, Cognitive, Perceptual, and Biological processes, as well as what Pennebaker calls

“Relativity”.

Social processes

This category contains a very large selection of 455 words and word stems in connection with human interactions, ranging from “ask” and “reply” to “partner” and “pal” and also including most personal pronouns and familial roles.87 In his baseline, Pennebaker finds that this particular subset of terms represents 8.32% of total English–language production. LIWC analysis of our corpora reveals that the shareholder letters originally written in English show a very strong correlation to this baseline, coming in at 8.95%, a fractionally higher usage rate for the words in this category, with an LSM of .96. The translated texts from BCV align almost perfectly between the average usage rate in the English–language letters and the baseline rate, at 8.61%. The translated texts, however, show a somewhat lower incidence of terms from the category, as they make up only 6.56% of the words in those texts.

87 Campbell et al., columns AB-AD.

57 Graph 16: Usage rates for social process terms

This disparity is minor enough that we would refrain from any attempt to ascribe a particular significance to it. Indeed, there is actually much higher variation between individual banks than the averages would suggest, with each language sub–group providing one of the highest concentrations of social–process words: Wells Fargo and Desjardins show usage rates of 12.33% and 11.36%, respectively, both significantly higher than the averages from our corpora as well as Pennebaker’s baseline. On the other hand, no English–language financial institution returns a usage rate of social–process words below 7.15% (HSBC’s showing), while seven out of ten of our French–language companies were below that mark. This suggests that the disparity, on a case by case basis, is in fact rather higher than it would appear from analysis of the average numbers, and may mean that the translation teams at some of our targeted institutions are employing various terms from this category less frequently than is the case for the writers who are crafting equivalent texts directly in English. However, the length of the word list composing this dictionary entry means that tracking the individual terms under–

represented in the translated texts is not a practical endeavor, at least from the starting point of this category. Fortunately, the linguistic–process categories covering personal pronouns have already shown what must be at least a portion of this disparity in greater relief and with more precision at the word level, and as such provide a much more practical call to action for financial translators than this more encompassing psychological–process category is able to do. At any rate, the LIWC numbers for our

0 1 2 3 4 5 6 7 8 9 10

Baseline English Translated BCV