Discussion on Learners’ Translation Performance

In the previous sections, the three groups of learners’ pretest and two posttests performance were analyzed and compared to explore the effectiveness of two web-based bilingual referencing tools on students. In addition, their perceptions toward the two bilingual referencing tools were presented. In this section, the findings in the present study will be discussed and compared with those reported in previous research.

Discussion on each research question is presented below.

The first and second research question aims to investigate whether the use of bilingual referencing tools could improve learners’ translation performance. The ANOVA results showed that the learners’ posttest scores of the translation task was significantly higher than their pretest scores, which reflects that the use of bilingual concordancer enables learners to select the contextually-appropriate expressions. This finding was consistent with those of Cheng & Gao (2016) and Tseng (2009): the use of bilingual concordancer enhanced the quality of the learners’ translation. Furthermore, a comparison between learners’ pretests and posttests showed that learners’

improvements were mostly in vocabulary or collocation. For example, most learners wrote “food danger” or “food risk” for “糧食危機.” And most learners usually did not have answers for “生態多樣性” and “探索奧秘.” After consulting the concordancer, they were able to obtain the correct vocabulary and thus gain points. However, their

89

syntactic errors were not corrected as much. It was observed that in the pretests, other than lexical errors, most of learners’ syntactic errors were related to the inflection form of a verb and sentence structures. It seemed that most learners were not aware of these errors or did not know how to correct them, even with a referencing tool at hand. Based on learners’ responses in questionnaires, the learners also considered the concordancer less helpful in terms of their grammatical problems. These findings were in correspondence with those of Tseng (2009) and Vannestål and Lindquist (2007).

To answer the third research question, all three groups of learners’ translation performance was assessed at three time points (pretest, posttest, and delayed posttest).

The posttest results showed that the experimental groups both significantly outperformed the control group, which suggested that the use of referencing tools helped learners to perform better on translation tasks. This finding is also in line with the previous studies that maintained that even novice translators, high school students in this case, could benefit from turning to referencing tools (Chen & Yu, 2013; Yepes, 2011). Additionally, the results also revealed that participants in Concordancer Alpha group progressed significantly more than Yahoo Dictionary group, which implied that bilingual concordancer was more effective in helping learners achieve better translation performance. The findings were similar to those in Wang’s research (2011), whose research focused on learners’ collocation learning with the help of referencing tools.

Based on learners’ responses to the questionnaire, they found the highlighted search word and target translation in the bilingual concordancer beneficial to their translation.

This allowed them to quickly locate their search word and target translation in the search results. In addition, the provision of copious example sentences and function of filtering example sentences by different translations helped them understand the differences in possible translations. These features of concordancer might have

90

contributed to learners’ significant progress in the posttests. In fact, it was the above-mentioned features that differentiate Concordancer Alpha from Yahoo Dictionary. To complete translation task where Chinese sentences should be translated into English, participants using Yahoo Dictionary would query in Chinese. However, when consulting in Chinese, the search results only contained its English equivalents without any example sentences or even indication of part of speech. This means that learners needed to query each possible translation to gain more information about these equivalents. However, it was probable that learners did not query further but randomly chose from the possible translations listed. One participant wrote in questionnaire that it would be better if each possible translation could show its frequency so that users could know how often the search word had been translated into these equivalent. In other words, this participant was aware that the showing of frequency might help learners determine which translation was most likely to fit the context. As suggested by Bowker (2003), although this word-count feature alone does not always guarantee accuracy of the word choice, it helped the users to determine which translation equivalent was more appropriate and more commonly used when leaners had more than one alternative L2 words. As a matter of fact, this feature is provided in Concordancer Alpha. And this might have helped participants in Concordancer Alpha Group choose more contextually-appropriate translation, thus improving their translation quality.

Another results yielded in ANOVA analysis was that although the two experimental groups progressed significantly in the posttests, their scores regressed significantly in the delayed posttests as well. The decline of the scores indicated that learners might have forgotten some of the learned expressions two weeks after the posttest. Several possible reasons might be able to account for the drop in the scores of delayed posttests.

91

First, participants were not given opportunity to practice new-learned expressions after the posttest. Due to lack of practice and repeated exposure of the new inputs, learners were unlikely to remember them after one exposure. Second, since learners were simply asked to write down their search results, they were not consciously memorizing the new words during their consultation process. Third, they were not given immediate feedback on their translation performance, so they were unsure if their translations were acceptable. Furthermore, there were other possible translations provided by the referencing tools, which might increase the uncertainty about the correctness of their translations. It’s possible that they did not try to memorize these new input when they couldn’t make sure whether to memorize them. These reasons together might have influenced learners’ performance in the posttest.

Nevertheless, the two experimental groups’ delayed posttest scores were still significantly higher than the control group’s delayed posttest scores and their own scores in pretests. Furthermore, the scores of Concordancer Alpha group were still significantly higher than those of Yahoo Dictionary Group. This indicates that the effects of referencing tools consultation could maintain for a period of time, which echoed with previous research (Chan & Liou, 2005; Yu, Liou, Chang, & Vongpumivitch, 2011): the residual effects of concordancing could be maintained over a period of time.

The findings also implied that bilingual concordancer helped learners to retain learned words and translations to such extent that their performance was still better than that in Yahoo Dictionary Group.