Accept/Reject/Revise: Distribution

For the FAQ text, the distribution of the matches into the three categories accept, reject and revise is shown in Table 7 for each translator and in Table 8 for all translators combined.

Chapter 5. Analyses and results 103

Match typeNo. of matches in text

Translator ATranslator BTranslator C Accept RejectReviseAccept RejectReviseAccept RejectRevi CM 2 2 100%0 0%0 0%2 100%0 0%0 0%2 100%0 0%0 100%282796%0 0%1 4%28100%0 0%0 0%2796%0 0%1 95-99%4 2 50%0 0%2 50%4 100%0 0%0 0%2 50%1 25%1 85-94% 6 0 0%3 50%3 50%0 0%3 50%3 50%0 0%4 67%2 75-84%3 0 0%0 0%3 100%1 33%0 0%2 67%0 0%1 33%2 70-74% 3 0 0%0 0%3 100%0 0%1 33%2 67%0 0%1 33%2 MT 303 10%0 0%2790%3 10%4 13%2377%2 7%6 20%22 In total 763445%3 4%3951%3850%8 11%3039%3343%1317%30 Match typeNo. of matches in text

Translator DTranslator ETranslator G Accept RejectReviseAccept RejectReviseAccept RejectRevi CM 2 2 100%0 0%0 0%2 100%0 0%0 0%2 100%0 0%0 100%282382%0 0%5 18%28100%0 0%0 0%2693%0 0%2 95-99%4 0 0%0 0%4 100%1 25%1 25%2 50%2 50%0 0%2 85-94% 6 0 0%3 50%3 50%0 0%4 67%2 33%0 0%3 50%3 75-84%3 0 0%0 0%3 100%0 0%0 0%3 100%0 0%0 0%3 70-74% 3 0 0%1 33%2 67%0 0%1 33%2 67%0 0%1 33%2 MT 305 17%7 23%1860%2 7%8 27%2067%2 7%1 3% 27 In total 763039%1114%3546%3343%1418%2938%3242%5 7%39

Chapter 5. Analyses and results 104

Match typeNo. of matches in text

Translator H Accept RejectRevise CM 2 2 100%0 0%0 0% 100%2825 89%0 0%3 11% 95-99%4 1 25%0 0%3 75% 85-94% 6 0 0%3 50%3 50% 75-84%3 0 0%0 0%3 100% 70-74% 3 0 0%1 33%2 67% MT 302 7%0 0%2893% In total 763039%4 5%4255% Table 7. Accept/reject/revise for each translator - FA Match typeTotal no. of matches in text

All translators Accept RejectRevise CM 1414100%0 0%0 0% 100%19618494%0 0%126% 95-99%281243%2 7%1450% 85-94% 420 0%2355%1945% 75-84%211 5%1 5%1990% 70-74% 210 0%6 29%1571% MT 210199%2612%16579% In total 53223043%5811%24446% Table 8. Accept/reject/revise for all translators - FAQ

Tables 7 and 8 show that the two CM matches were accepted by all translators, which was also to be expected since these are 100% matches that occur in the same context as the segment stored in the TM. As shown in Table 8, the 28 100% matches were generally accepted (94%) by the translators. The remaining 100% matches were revised (6%), i.e. no 100% matches were rejected. Table 7 shows that Translators B and E accepted all 28 (100%), Translators A and C accepted 27 (96%) and Translators G, H and D accepted 26 (93%), 25 (89%) and 23 (82%), respectively. The four 95-99% matches were mostly revised (50%) or accepted (43%) and on a few occasions rejected (7%). Table 7 shows that the individual translators’ choices were quite different here; for example, Translator B accepted all four 95-99% matches, whereas Translator D revised all four. The 6 matches in the 85-94% match category were mostly rejected by the translators (55%). Considering that all other TM match types were far less frequently rejected, this was a bit surprising. However, an explanation can be found in the experimental setup, as will be explained in Section 5.2.2.2.2. The remaining 85-94% matches were revised (45%), i.e. no 85-94% matches were accepted. The individual translators’ choices were quite similar here: of the six matches within this match type, Translators A, B, D, G and H all rejected three and revised three, and Translators C and E rejected four and revised two of the matches. The three 75-84% matches were mostly revised (90%) by the translators. Translator B accepted one 75-84% match and Translator C rejected one, but apart from that, all 75-84% were revised. The translators also mostly revised the three 70-74% matches (71%). The matches that were not revised were rejected (29%). All translators except Translator A (who revised all three matches) revised two of the 70-74% matches and rejected one. The 30 MT matches were mostly revised (79%). The remaining MT matches were rejected (12%) or accepted (9%). Thus, in 9% of the cases where the translators were offered an MT match, they accepted it without modifications.

Differences in the translators’ choices were observed here. Translators C, E, G and H each accepted two MT matches, Translators A and B accepted three and Translator D accepted five matches. Translators A and H did not reject any MT matches, whereas Translators G, B, C, D and E rejected one, four, six, seven and eight MT matches, respectively.

Figure 14 shows the percentages of matches that were accepted, rejected and revised in each match type and by all translators in total. From the figure it is clear that fuzzy matches with higher match values were the match types that were most frequently accepted, and that fuzzy matches with lower match values and MT matches were the match types that were most frequently revised. Apart from the high percentage of 85-94% matches that were rejected (as noted above), 70-74% and MT matches were the match types most frequently rejected.

Figure 14. Accept/reject/revise for match types and all translators - FAQ text Regarding the TM/MT threshold, it is interesting to observe that the translators rejected 29% of the 70-74% matches compared to 12% of the MT matches. Since intuitively we would expect translators to reject matches of a low quality, this might indicate that the translators find that the 70-74% are of a lower quality and thus less worth revising than MT matches.

However, the amount of data in the 70-74% category is quite low, and, as will be explored further in Section 5.2.2.2.2, the 6 rejected 70-74% matches involve six translators rejecting the same 70% match.

For the Newsletter, Tables 9 and 10 show the distribution of the matches into the categories accept, reject and revise for each translator and for all translators, respectively. As shown in Table 9, all translators but one (Translator H) accepted the one 100% match (100%), and the one 95-99% match was revised by all translators (100%). The two 85-94% matches were mostly revised (79%) by the translators, although three translators (B, G and H) each accepted one of the matches (21%). Six of the seven translators revised all five 75-84%

matches (94%), and Translator G rejected two of the five matches, which equals 6% of the total number of 75-84% matches. The three 70-74% matches were revised by six of the seven translators (95%). Translator C accepted one 70-74% match, which equals 5% of the total number of matches in this match type. The translators mostly revised the 13 MT matches in the Newsletter (93%). Two MT matches were rejected in total (2%) (Translators B and E each rejected one MT match) and four were accepted (4%) (Translators E and G each accepted two MT matches). Thus, in 4% of the cases where the translators were offered an MT match in the Newsletter, they accepted it without modifications.

0 10 20 30 40 50 60 70 80 90 100

CM 100% 95-99% 85-94% 75-84% 70-74% MT

Accept Revise Reject

Chapter 5. Analyses and results 107

Match typeNo. of matches in text

Translator ATranslator BTranslator C Accept RejectReviseAccept RejectReviseAccept RejectRevi 100%1 1 100%0 0%0 0%1 100%0 0%0 0%1 100%0 0%0 95-99%1 0 0%0 0%1 100%0 0%0 0%1 100%0 0%0 0%1 85-94% 2 0 0%0 0%2 100%1 50%0 0%1 50%0 0%0 0%2 75-84%5 0 0%0 0%5 100%0 0%0 0%5 100%0 0%0 0%5 70-74% 3 0 0%0 0%3 100%0 0%0 0%3 100%1 33%0 0%2 MT 130 0%0 0%13100%0 0%1 8%1292%0 0%0 0%13 In total 251 4%0 0%2496%2 8%1 4%2288%2 8%0 0%23 Match typeNo. of matches in text

Translator DTranslator ETranslator G Accept RejectReviseAccept RejectReviseAccept RejectRevi 100%1 1 100%0 0%0 0 1 100%0 0%0 0%1 100%0 0%0 95-99%1 0 0%0 0%1 100%0 0%0 0%1 100%0 0%0 0%1 85-94% 2 0 0%0 0%2 100%0 0%0 0%2 100%1 50%0 0%1 75-84%5 0 0%0 0%5 100%0 0%0 0%5 100%0 0%2 40%3 70-74% 3 0 0%0 0%3 100%0 0%0 0%3 100%0 0%0 0%3 MT 130 0%0 0%13100%2 15%1 8%1077%2 15%0 0%11 In total 251 4%0 0%2496%3 12%1 4%2184%4 16%2 8%19

Chapter 5. Analyses and results 108

Match typeNo. of matches in text

Translator H Accept RejectRevise 100%1 0 0%0 0%1 100% 95-99%1 0 0%0 0%1 100% 85-94% 2 1 50%0 0%1 50% 75-84%5 0 0%0 0%5 100% 70-74% 3 0 0%0 0%3 100% MT 130 0%0 0%13100% In total 251 4%0 0%2496% Table 9. Accept/reject/revise for each translator - New Match typeTotal no. of matches in text

All translators Accept RejectRevise 100%7 6 86%0 0%1 14% 95-99%7 0 0%0 0%7 100% 85-94% 143 21%0 0%1179% 75-84%350 0%2 6%3394% 70-74% 211 5%0 0%2095% MT 914 4%2 2%8593% In total 175148%4 2%15790% Table 10. Accept/reject/revise for all translators - New

Figure 15 shows the percentages of matches which were accepted, rejected and revised in each match type by all translators in total. From the figure, we see that 100% matches were mostly accepted, whereas all other match types were mostly revised. 75-84% and MT matches were the only match types which were rejected.

Figure 15. Accept/reject/revise for match types and all translators - Newsletter

Regarding the TM/MT threshold, there are only small differences between the translators’

choices in 70-74% and MT matches (and 75-84%, for that matter). When only seen from the viewpoint of the categorization into accept, reject and revise, this might indicate that these fuzzy match types and MT matches are comparable. However, this needs to be seen in the light of other results such as editing speed in these match types.