Chapter 3 Developing a Revision Scheme that Accounts for Both TTs and STs
3.3 Mapping of the ATA and DQF-MQM error schemes
Given the standing and complementarity of the ATA and DQF-MQM error schemes, the first step is to map these two selected typologies into a merged scheme, removing error types that do not apply to JA to ZH-CN translations of hybrid texts and
67
merging similar error types to a single category. Table 3.1 shows the mapping of the ATA and DQF-MQM schemes.
The [CSR scheme (TT)] column in Table 3.1 shows the initial list of categories adopted for establishing the revision scheme—the CSR scheme—customized for assessing ZH-CN translations of JA CSR reports. Since this CSR scheme aims to include overt problematic source features that may inhibit translation quality, it prefixes error categories with “TT” to indicate translation errors in the TT. The [ATA category label] column and [ATA category label] column show error categories that exist in their schemes, respectively.
The TT error categories fall into six clusters. A discussion of specific cases follows.
No. 1 to No. 8 are error categories that the CSR scheme adopted from both ATA and DQF-MQM where there is an equivalence established between ATA and DQF-MQM for each error category.
No. 9 to No. 15 are error categories that the CSR scheme adopted from ATA, which are absent from DQF-MQM.
No. 16 to No. 21 are error categories that the CSR scheme adopted from DQF-MQM, which are absent from ATA.
No. 22 to No. 25 are original categories added to the CSR scheme.
No. 26 to No. 29 are error categories that the CSR scheme did not adopt from ATA. These categories exist in ATA but are absent from DQF-MQM.
No. 30 to No. 35 are error categories that the CSR scheme did not adopt from DQF-MQM. These categories exist in DQF-MQM but are absent from ATA.
68
Table 3.1 Mapping of ATA and DQF-MQM
No. ATA category label CSR scheme (TT) DQF-MQM category label 1 Addition TT-Accuracy-Addition Accuracy-Addition 2 Omission TT-Accuracy-Omission Accuracy-Omission 3 Mistranslation
Misunderstanding
TT-Accuracy-Misrepresentation Accuracy-Mistranslation 4 Punctuation
TT-Fluency-Punctuation Fluency-Punctuation 5 Spelling/Character
Capitalization (a sub-category of Spelling/Character) Diacritical
marks/Accents (sub-category of Spelling/Character)
TT-Fluency-Spelling/Character Fluency-Spelling
6 Grammar Syntax
(a sub-category of Grammar)
Word form/Part of speech
(a sub-category of Grammar)
TT-Fluency-Grammar Fluency-Grammar
7 Usage
TT-Fluency-Awkward/Unidiomatic Style-Awkward Style-Unidiomatic 8 Other Errors TT-Other errors Other
9 Faithfulness TT-Accuracy-Faithfulness
10 Faux ami TT-Accuracy-False friend
11 Ambiguity TT-Accuracy-Ambiguity 12 Unfinished
TT-Accuracy-Unfinished 13 Literalness
TT-Accuracy-Literalness 14 Text Type (includes the
former categories of Register and Style)
TT-Fluency-Register 15 Cohesion TT-Cohesion
16
TT-Accuracy-Over-translation Accuracy-Over-translation 17
TT-Accuracy-Under-translation Accuracy-Under-translation
18
TT-Accuracy-Untranslated Accuracy-Untranslated
19
TT-Fluency-Inconsistency Fluency-Inconsistency
20 TT-Design Design-Local formatting
Design-Markup Design-Length
Design-Truncation/text expansion
Design-Missing text 21 TT-Locale convention Locale
convention-Address format
69
Date format Currency format Measurement format Shortcut key
Telephone format
22
TT-Accuracy-Transliteration 23 TT-Accuracy-Calque
24 TT-Terminology
25 TT-TM
26 Illegibility 27 Indecision 28 Terminology 29 Verb Tense
30 Accuracy-Improper exact
TM match
31 Fluency-Grammatical
register
32
Fluency-Link/cross-reference
33
Terminology-Inconsistent with termbase
Terminology-Inconsistent use of terminology
34 Style-Inconsistent style
Style-Company style Style-Third-party style
35 Verity-Culture-specific
reference
For error categories No. 1 to No. 8, an equivalence can be established between ATA and DQF-MQM. For instance, [TT-Accuracy-Addition] was adopted from [Addition] in the ATA scheme, which equates to [Accuracy-Addition] in the DQF-MQM scheme. For No. 3, [TT-Accuracy-Misrepresentation] in the CSR scheme replaced “mistranslation” (ATA and DQF-MQM) and “misunderstanding” (ATA) with
“misrepresentation” to emphasize the importance of making judgments based on textual evidence, because “misunderstanding” implies that assessors need to guess what was going on in the translator’s mind. Also, “mistranslation” is too broad to be regarded as a subcategory. For simplification, No. 5 was merged with Spelling/Character], No. 6 with Grammar], and No. 7 with [TT-Fluency-Awkward/Unidiomatic].
70
For categories adopted from the ATA scheme, [Faithfulness], [Faux ami]
(renamed to [False friend] for better understandability), [Ambiguity], [Unfinished], [Literalness], and [Register] (taken from the ATA’s former category) are—based on the present author’s personal experience—error categories that are likely to be observed in professional JA to ZH-CN translations. On the same grounds, [Accuracy-Over-translation], [Accuracy-Under-[Accuracy-Over-translation], [Accuracy-Untranslated], [Fluency-Inconsistency], [Design] (all categories merged), and [Locale convention] (all sub-categories merged) were adopted from DQF-MQM.
Regarding original TT error categories added to the CSR scheme, firstly, [TT-Accuracy-Transliteration] is an error type unique to JA to ZH-CN translation, where a JA word is converted to the simplified version of the same kanji in ZH-CN, even though it does not, in fact, exist in ZH-CN. For example, “介護” (nursing-care) is translated as
“介护” using the simplified version of “介護,” even though “介护” does not exist in ZH-CN.
Secondly, [TT-Accuracy-Calque] is another error type often observed in professional JA to ZH-CN translation, translating the whole sentence word for word inheriting the structure of the ST. As shown in Example 3-1, the EN gloss of the JA text and that of the published ZH-CN text is almost the same, indicating that the published ZH-CN text followed the wording and the word order as is in JA. It is considered as a bad ZH-CN translation because: (1) “向 N1 提供 N2” (providing N2 to N1) is a fixed structure in ZH-CN. The translation reads unnaturally if the “向 N1” (to N1) part is omitted as the meaning will become ambiguous; (2) it is necessary to add the auxiliary verb “能” (can) in front of the verb “支援” (support). The verb “支える” (support) in JA is in its present form, but ZH-CN translation needs to supplement the auxiliary verb as necessary as is required in EN; and (3) it is unnatural to use “未来时代” (future era) in ZH-CN which is a word-for-word translation of “これからの時代” (korekarano jidai). The whole sentence would be labeled as [TT-Accuracy-Calque] because the
ZH-71
CN text is a word-for-word translation, following the exact sentence structure in JA, ignoring the proper tense of verbs and the natural wording.
Example 3-1 Example of [TT-Accuracy-Calque]
JA XXXでは、これからの時代の医療を支える新し いソリューションを提供しています。
EN gloss of JA At XXX, offering new solutions that support health care of future era.
Published ZH-CN XXX提供着支援未来时代医疗的全新解决方案。
EN gloss of ZH-CN XXX is providing new solutions that support future era healthcare.
Example 3-2 is another example that explains [TT-Accuracy-Calque]. The JA text, a noun cluster in this example, is a proper noun created by the company that published the CSR report. The intended meaning is “restricted substance that contains TIG.” Both “含有” (contain) and “制限” (restrict) are sahen-nouns in JA, and they become verbs by adding the stem-verb suru. When they are connected with other nouns to form a noun cluster, it is difficult to tell if the voice is active or passive because the element, suru or sareru, that is used to indicate the voice has been omitted. Readers have to decide the relation between words based on the context. As shown in Example 3-2, there are two possible EN glosses, and the grayed-out version is the wrong interpretation; the ZH-CN translation adopted the kanji words in JA as is and followed the exact word order, which yields “TIG contains restricted substance,” the opposite of the intended relation. Therefore, the whole sentence would be labeled as [TT-Accuracy-Calque] as it is an accuracy error, and what caused the error is that the translation borrowed the phrase from JA by word-for-word translation.
72
Example 3-2 Example of [TT-Accuracy-Calque]
JA 「TIG含有制限物質」
EN gloss of JA Gloss 1: “TIG contain restrict substance”
Gloss 2: “restrict substance containing TIG”
Published ZH-CN “TIG含有限制物质”
EN gloss of ZH-CN “TIG contain restrict substance”
Thirdly, [TT-Terminology] is a combination of terminology related categories in ATA and DQF-MQM. By ATA’s definition, a terminology error occurs when a term appropriate to a specific subject field is not used when the corresponding term is used in the ST. DQF-MQM, on the other hand, defines terminology error as (a) violation of specified glossary and (b) inconsistency in translating the same term. [TT-Terminology]
in the CSR scheme covers all three cases.
Fourthly, [TT-TM] in the CSR scheme indicates errors such as where a translator has applied the wrong translation memory (TM) or translated a sentence without referring to a 100% match in the TM, which may violate the client’s instruction depending on the agreement between clients and LSPs on how to work on 100%
matches. [TT-TM] also includes the TM-related category (No. 30) in DQF-MQM
“Improper exact TM match” which indicates errors such as when a translator fails to amend a TT segment that is proposed on the basis of a 100% match (a match where there is no difference between the ST in the document and the ST in the TM) when the suggested translation does not, in fact, apply to the current context.
From the ATA scheme, the CSR scheme did not adopt [Illegibility], [Indecision], [Terminology], [Text Type] (includes the former categories [Register] and [Style]), and [Verb Tense] for the following reasons.
73
An [Illegibility] error applies only to handwritten texts in the setting of a certification exam when graders cannot read what the candidate has written.
An [Indecision] error occurs when the candidate gives more than one option for a given translation unit, which rarely occurs in professional settings.
[Terminology] has been adopted in a broader sense.
[Register], a sub-category of [Text Type], was adopted in the CSR scheme as [TT-Fluency-Register], while [Style], the other sub-category of [Text Type] (e.g., tone, method of exposition) was also subsumed under [TT-Fluency-Register].
[Verb Tense] including modality is considered a grammar error in the CSR scheme.
From the DQF-MQM scheme, the CSR scheme did not adopt the following categories for the following reasons.
[Accuracy-Improper exact TM match] has been adopted in a broader sense.
[Fluency-Grammatical register] is subsumed under [TT-Fluency-Register].
[Fluency-Link/cross-reference] is considered as a pre-processing or post-processing issue rather than a translation error. If the translator has translated the tags wrongly, it is deemed to be a localization-specific error which will be categorized as [TT-Design] under the CSR scheme.
[Terminology-Inconsistent with termbase] and [Terminology-Inconsistent use of terminology] are subsumed under [TT-Terminology].
[Style-Inconsistent style], [Style-Company style], and [Style-Third-party style] are subsumed under [TT-Fluency-Inconsistency].
[Verity-Culture-specific reference] should be regarded as a deficiency in the ST rather than a translation error.
74
As a result of mapping the ATA and DQF-MQM schemes, the CSR scheme has a list of TT error categories. Table 3.2 shows the list of categories created to indicate problematic source features in the TT, prefixing each feature with “ST.”
Table 3.2 ST categories in the CSR scheme CSR scheme (ST) ST-Incomprehensibility
ST-Ambiguity ST-Incompleteness ST-Grammar ST-Register
ST-Spelling/Character ST-Punctuation
ST-Awkward/Unidiomatic ST-Inconsistency
ST-Cohesion ST-Falsity ST-Complexity ST-Borrowed word ST-Illogicality ST-Word order ST-Other errors
Section 3.4 describes the process of validating the CSR scheme (a combination of TT categories in Table 3.1 and ST categories in Table 3.2) through the achievement of satisfactory inter-assessor agreement.