The reject category

As stated in Section 5.2.1, a match is categorized as belonging to the reject category if it is deleted without having been modified beforehand. This includes matches which were deleted because the translators used the Copy Source to Target function, by means of which the provided match is deleted and the source segment is inserted in the target segment instead. It also includes TM matches which are deleted because translators choose to replace them with an MT suggestion (cf. Section 4.3.1.1.3.1.2).

Table 18 shows that, in the FAQ text, all rejected matches except for one were rejected by means of the translators’ use of the Copy Source to Target function. The remaining match was rejected because the translator chose to insert the MT match instead of the proposed TM match. Interestingly, no matches were deleted by “pure” deletion, i.e. where translators deleted the match letter by letter or used the mouse or keyboard to select the match and delete it altogether. This is interesting, because translators are generally expected to translate source segments from scratch if a match is rejected. The present study shows that the translators did not, in fact, translate from scratch when they rejected a match; instead they replaced the match with either the source segment or an MT match and worked with this instead.

Match type Total number of rejected matches

Copy Source to Target Insertion of MT match "Pure" deletion

CM 0 0 0 0

Table 18. Rejection type for all translators - FAQ text

The two 95-99% rejected matches in the FAQ text include one match that was rejected by two translators, shown in Table 19.

Segment

number Match

type Source segment Match Rejected by

translator(s)

12 99% BeoVision 7-40/55 BeoVision 7-40/55.

BT: BeoVision 7-40/55. C, E

Table 19. Rejected 95-99% matches - FAQ text

Translators C and E both used the Copy Source to Target function in segment 12. The match contained a full stop which was not part of the source text, and since the translators

generally followed the layout of the source text, the translators probably wanted to delete the full stop. However, they may have found it easier to use the Copy Source to Target function here instead of deleting the full stop themselves, since the text in the source segment could be transferred directly to the target text segment and be used as the translation as well.

The 23 rejected 85-94% matches are visible from Table 20.

Segment

Table 20. Rejected 85-94% matches - FAQ text

As mentioned in Section 5.2.2.1, it seemed surprising that more than half of the matches in the 85-94% match category were rejected, considering that all other TM match types were far less frequently rejected. However, an explanation might be found in the experimental setup. As described in Section 4.3.1.1.3.1.4.1.2, words which were to remain untranslated in the target text were formatted in red in the source text, and the source text happened to contain three segments with a 94% match value which each contained only one word

marked in red (segments 26, 29 and 32). All seven translators chose to reject these three matches by using the Copy Source to Target function, because the source-text segment could be transferred directly to the target segment and be used as the translation. This accounts for 21 of the 23 rejected 85-94% matches, and the experimental setup therefore appears to have impacted on the results in an unfortunate way in this regard. In the last segment from the 85-94% match category which was rejected, segment 46, “BeoLab 14”

was marked in red in the source text, and the match contained a wrong product name (Beolit 14). This may have been the reason why Translators C and E chose to use the Copy Source to Target function in this segment, since in this way, the correct product name was transferred to the target segment together with the rest of the source text segment.

The one rejected 75-84% match was the match in segment 71, rejected by Translator C and shown in Table 21.

Technical Sound Guide. Du finder flere oplysninger i hurtigvejledningen.

BT: You find more information in the quick guide.

C

Table 21. Rejected 75-84% match - FAQ text

“Technical Sound Guide” was marked in red in the source text, and in line with the examples above, this might have been the reason why Translator C used the Copy Source to Target function to reject the match.

The six rejected matches in the 70-74% match category include one 70% match which was rejected by six of the translators, namely the match provided in segment 51 (cf. Table 22).

Translators B, C, D, E and H used the Copy Source to Target function in the segment, and Translator G chose to insert the MT match instead of the proposed TM match.

Segment

number Match

type Source segment Match MT match inserted

instead of TM match Rejected by translator(s) 51 70% [tag]A software update in

progress: En optagelse er i gang BT: A recording is in progress

B, C, D, E, H

51 70% [tag]A software update in

progress: En optagelse er i gang BT: A recording is in

Table 22. Rejected 70-74% matches - FAQ text

The source segment in segment 51 contained a tag indicating the presence of a visual element which was not present in the match. By using the Copy Source to Target function, the tag was transferred to the target segment, which may have motivated Translators B, C, D, E and H to use this function. After using the Copy Source to Target function, the

translators edited the source segment (which was now in the target segment) so that it was changed to a Danish translation. Interestingly, after Translator G inserted the MT match (which like the 70% TM match also did not contain the tag) into the target segment, he copied “En softwareopdatering i gang” (A software update in progress) in the MT match,

used Copy Source to Target and replaced “A software update in progress” with “En

softwareopdatering i gang”. Thus, Translator G also used the Copy Source to Target function in this segment to include the tag in the final translation.

The 26 rejected MT matches were all rejected by use of the Copy Source to Target function.

The matches are shown in Table 23.

Segment

number Match

type Source segment Match Rejected by

translator(s)

3 MT Which MODE setting to use

depends on the type of your Bang

& Olufsen television.

Hvilken TILSTAND indstilling afhænger af dit Bang & Olufsen fjernsyn.

BT: Which MODE setting depends on your Bang & Olufsen television. positionsknappen ( FRI , WALL , HJØRNE ) på BeoLab 14 ?

switch to standby after 3 minutes with no sound, when the MODE switch is set to LINE or AMP?

Hvorfor bliver min BeoLab 14 ikke omskifter til standby efter 3 minutter uden lyd, når tilstanden omskifter er indstillet til Line eller AMP ? BT: Why does my BeoLab 14 not become switch to standby after 3 minutes without sound, when the mode switch is adjusted to Line or AMP

?

E

41 MT To avoid noise on the LINE or AMP signal m ake sure that you have connected all sockets on BeoLab 14 to the corresponding sockets on the connected product.

For at undgå støj på Line eller AMP signal m ake sikker på, at du har tilsluttet alle stik på BeoLab 14 til de tilsvarende stik på det tilsluttede produkt.

BT: To avoid noise on Line or AMP signal m ake sure that you have connected all sockets on BeoLab 14 to the corresponding sockets on the connected produkt.

E

43 MT MODE switch set to AMP TILSTAND omskifter indstillet til AMP BT: MODE switch set to AMP E 44 MT MODE switch set to LINE TILSTAND omskifter indstillet til Line

BT: MODE switch set to Line E 52 MT Do not disconnect BeoLab 14 from

the mains during the software update.

Afbryd ikke BeoLab 14 fra lysnet under softwareopdatering.

BT: Do not diconnect BeoLab 14 from the mains during software update.

C, D, E

55 MT [tag]Switched on (10 seconds

duration) Tændt (10 sekunder varighed)

BT: Switched on (10 second duration) B, C, D 58 MT [tag]Switched off (10 seconds

duration) Slukket (10 sekunder varighed)

BT: Switched off (10 second duration) B, C, D, E

60 MT [tag]Product failure: Produktfejl:

BT: Product failure: C, D, E

63 MT [tag]Error indication, e.g. at

overheating: Fejl, f.eks. ved overophedning:

BT: Error, e.g. at overheating: C, D

Table 23. Rejected MT matches - FAQ text

All of the rejected MT matches had at least one of the following three characteristics: 1) the entire source segment text could be transferred directly to the target text segment and used as the translation in an unchanged form (segment 5), 2) the source segment contained one or more red words which should remain untranslated in the target text (such as BeoLab 14,

MODE, FREE, WALL, CORNER, LINE, AMP and Technical Sound Guide, cf. Section 4.3.1.1.3.1.4.1.2) (segments 3, 23, 39, 41, 43, 44, and 52), and 3) the source segment contained a tag indicating the presence of a visual element in the text which also had to be present in the target segment (segments 55, 58, 60 and 63). In the first case, in Schjoldager’s (2010) terminology, the microstrategy of “direct transfer” is possible. Schjoldager (2010, p.93) notes herself that copying the source text to the target segment is an instance of direct transfer. Concerning the second and third cases, the data suggest that the translators

(primarily Translators B, C, D and E) preferred rejecting the match and editing the source segment instead of the match when the source segment contained red words that should remain untranslated or visual elements (this is further addressed in Section 5.2.2.2.3.3.2). In these cases, parts of the source segments could be used in the translation in an unchanged form (the red words and the tags), and the translators needed to translate the remaining parts of the segments from English into Danish. This is interesting since the translators could also, for example, have copied the red words and tags from the source segment and pasted them into the target segment. In this case, the match would have remained in the target segment and could have been used by the translators. In the retrospective interviews, all translators except Translator A explained that they use the Copy Source to Target function when the source text contains tags, formatting, names and/or words that should remain untranslated in the target text. In the experiment, words were only marked by red to make translators aware that they should remain untranslated in the target text (and not because they should also be marked in red in the target text), but the translators generally made sure that the words were also marked in red in the target text. Thus, in the case of the red words, we cannot know whether the translators used the Copy Source to Target function to transfer the red formatting to the target text or to avoid typing the words in the target text

themselves (which would also involve the risk of misspelling) or both. Put differently, we cannot know whether the translators would have typed the words in the target text

themselves if they had not been formatted in red or whether they would also have used the Copy Source to Target function.

Interestingly, all of the rejected TM matches addressed above also had one or more of these three characteristics. In segments 12, 26, 29, 32, direct transfer of the whole source segment was possible, segments 26, 29, 32, 46 and 71 contained red words and segment 51

contained a tag indicating the presence of a visual element. Thus, the analysis indicates that certain elements trigger the rejection of MT matches, first and foremost by use of the Copy Source to Target function.

Finally, in relation to source segments that contained a visual element, it is interesting to observe that the visual elements had not been included in the matches. From a TCI perspective, it is interesting to consider whether the translators would have revised, not rejected, the matches if the visual elements had been included in the matches.

Unfortunately, I have no knowledge of the mechanisms behind this.

In the Newsletter, four matches were rejected. As shown in Table 24, two were rejected using the Copy Source to Target function and two by insertion of MT match instead of the proposed TM match. No matches were rejected by “pure” deletion, i.e. none of the

translators rejected a match by, so to speak, “manually” deleting each letter or using the mouse or keyboard to select the match and delete it.

Match type Total number of rejected matches

Copy Source to Target Insertion of MT match "Pure" deletion

100% 0 0 0 0

Table 24. Rejection type for all translators - Newsletter

The two rejected 75-84% matches are shown in Table 25. These were both rejected by Translator G and, in both cases, he replaced the TM matches with the provided MT matches.

Segment

number Match

type Source segment Match MT match inserted

instead of TM match Rejected by translator(s)

2 77% BeoSound 5 relaunched

with Spotify integration BeoSound 5 – med Spotify Inside

Table 25. Rejected 75-84% matches - Newsletter

In segment 2, the MT match provided is a more adequate translation of the source segment than the 77% match pre-inserted into the target segment which is probably the reason why Translator G replaced the TM match with the MT match. After insertion of the MT match, Translator G replaced the hyphen between “5” and “relanceret” with a dash and deleted integration”. Later in the editing phase, the translator returned to the segment and wrote “-integration” after “Spotify”, thus reproducing what was suggested by the MT engine. In segment 9, although “lydsystem” (sound system) is not a correct rendering of “retail concept”, the MT match provided seems to be closer to an adequate translation of the source segment than the pre-inserted 78% match which probably made Translator G replace the TM match with the MT match. Translator G’s rejection of segment 9 is illustrated below (example 7-NL-G-9).

The two rejected MT matches include one match which was rejected by two translators (B and E) (cf. Table 26). Both translators rejected the match using the Copy Source to Target function.

Segment

number Match

type Source segment Match Rejected by

translator(s) 22 MT B&O PLAY television wins coveted

red dot award B&O afspille tv vinder den eftertragtede rød prik pris BT: B&O play tv wins the coveted red dot award

B, E

Table 26. Rejected MT matches - Newsletter

Using the Copy Source to Target function might have been motivated by “B&O PLAY” and

“red dot award”, which both translators retained in the final translation. Translator E’s process in segment 22 is illustrated in the process examples below (example 8-NL-E-22).

It is somewhat surprising that only four matches were rejected (equalling 2%) in the Newsletter, particularly as a couple of MT matches seemed to be of quite a low quality, as noted by most of the translators in the retrospective interviews. Especially the matches in segments 11 and 12 (see Table 27) were said to be almost gibberish or nonsense, with Translator H even stating that the match in segment 11 is an example of MT “at its worst”.²⁷ However, all of the translators chose to revise these matches instead of rejecting them. For some of the translators (B, D and F), this may be explained by the information given in the webinar on post-editing which they attended approximately three months prior to the experimental study (cf. Section 5.1.1), where they were told not to “ignore or erase the machine translation output and translate from scratch” and to “[r]etain as much raw output as possible”. Translator B mentioned something similar in the retrospective interview when he said that the recommendation is to revise the MT output instead of just discarding it, although without referring to the webinar. Also, as mentioned in Section 4.3.2.1, since the study was conducted around the time when MT was first implemented at TextMinded, the translators may have expected me to be particularly interested in MT and thus may have thought that they were required to work with the MT matches proposed to them, even though I never specifically told them so. However, the translators did not seem to follow this in the translation of the FAQ text, where a higher percentage of MT matches was rejected (12%, cf. Section 5.2.2.1).

Segment

number Match

type Source segment Match

11 MT Experiencing the brand’s acoustic innovations first-hand has never been more accessible or compelling.

Opleve den brand akustiske nyskabelser det har aldrig været nemmere lydfronten.

BT: Experience that brand acoustic innovations it has never been easier the sound front.

12 MT In an atmosphere that is luxurious but welcoming, the new design captivates curiosity as you move through the store.

I en atmosfære, luksuriøst, men imødekommende, det nye design samtidig når man går gennem gemme.

BT: In an atmosphere, luxurious, but welcoming, the new design simultaneously when you walk through hide.

Table 27. Examples of MT matches - Newsletter

27 Cf. also the HTER analysis in Section 5.6, where segments 11 and 12 were among the segments with the highest HTER scores, i.e. among the segments in which most edits were made.

5.2.2.2.2.1 Process examples

In the following, two process examples from both the FAQ text and the Newsletter are provided for the reject category.

Example 5-FAQ-D-3:

One of the matches that was rejected using the Copy Source to Target function in the FAQ text is the match in segment 3, an MT match rejected by Translator D. The translator started out by using the Copy Source to Target function, replacing the match with the source text segment (step 2). During the retrospective interview, the translator stated that the reasons she used the Copy Source to Target function were that the match had to be reformulated in its entirety, that something from the source segment had to be retained in the target text (“MODE” which was formatted in red and had been translated into “TILSTAND” by the MT engine) and that the match contained tags that were not in the right order. Thus, she would rather copy the source segment to the target segment instead of moving these elements around. In steps 3 and 4, the translator alternated between deleting parts of the English source text and typing the Danish translation. When typing “indstillingen” (step 3), the translator was offered “indstilling” by the AutoSuggest function, but she did not accept the suggestion, although it was very close to what she ended up writing. During the interview, I asked the translator if she remembered whether she drew on the suggested match (which was still visible to her in the Translation Results Window in the upper part of the screen) after using the Copy Source to Target function. She answered that she translated based on the source segment, meaning that she did not draw on the MT match. Also, she mentioned that she was uncertain whether “television” should be translated as “fjernsyn” (television) or

“tv” (TV) and that she made a quick decision and used “fjernsyn”, thinking that she might

“become wiser” later in the text, i.e. the subsequent text may confirm her decision or cause her to change her mind.

Steps Process description Interim target segment Interview

1 Wording of provided match Hvilken TILSTAND[tag][tag] indstilling afhænger af dit Bang & Olufsen fjernsyn.

BT: Which MODE setting depends on your Bang & Olufsen television.

2 Uses Copy Source to Target. Which [tag]MODE[tag] setting to use depends on the type of your Bang &

Olufsen television.

App. 6, D, ll. 235-246 3 Selects "Which" and deletes it.

Writes "-indstillingen" after

"MODE" and the tag. While writing

"-indstillingen", the translator is offered "indstilling" by the AutoSuggest function. It is not used by the translator.

[tag]MODE[tag]-indstillingen setting to use depends on the type of your Bang &

Olufsen television.

App. 6, D, ll. 249-263

4 Writes "afhænger af dit". Deletes

"setting to use depends on the type of your". Writes "-fjernsyn" after

"Olufsen". Deletes "television".

[tag]MODE[tag]-indstillingen afhænger af dit Bang & Olufsen-fjernsyn.

During the checking phase, the translator entered segment 3 and implemented changes

During the checking phase, the translator entered segment 3 and implemented changes

In document Aalborg Universitet (Post-)Editing - A Workplace Study of Translator-Computer Interaction at TextMinded Danmark A/S Bundgaard, Kristine (Sider 131-143)