Test & Comparison of Google, Baidu, Bing on Chinese Translation
Google translate is on the top of ranks for free translation apps. Google Neural Machine Translation (GNMT), its self-developed translation engine, has revealed Google's almighty power and astonishing learning ability in this field. Nevertheless, with its dropping out of China mainland in 2010, it may suffer from weak connections or deferred collection of word sources.
Baidu, Chinese local Internet tycoon, Google's biggest counterpart in China, started its online translation service dozens years ago. Without tremendously innovative technologies, Baidu translation, however, established its strong corpus through unlimited and massive language source accumulation. Bing translation, based on Microsoft, has also been well trained through lots of user experiences and machine learning.
2019 is a great year for machine learning. But here we are not going to discuss further about underlying principles or theories about it. As what normal users concerns about is their effectiveness. So we have chosen a series of examples from different topics and put them into these 3 translation web apps to check and compare their correctness and usability.
1. Daily Dialogs
This is a very basic part of human communications, as daily conversations have the greatest frequency and function of utilization. At the same time, all translation apps providers have the best access to a large number of vocabularies and thus can easily perform the tasks. Below is our test:
- 早上好！ (Good morning!)
- 你好！ (Hello!)
- 昨晚的电影怎么样？ (How was the movie last night?)
- 挺好的。情节还行。另外，演员我也比较喜欢，挺有趣的。 (It was terrific. The plot was fine. Meanwhile, I like the actor, who was very funny.)
From the above test, 3 results are all acceptable. For that long sentence, there are slight differences between each results or flaws on grammar, but anyway, they are making sense. Meanwhile, there are line breaking disorders for Bing translation.
2. Academic Paper (Basic)
Words and phrases used in an academic thesis could be formal, precise, and professional. Failing to translate terminologies or conventional expressions appropriately for such document often means big failure. Here is our test:
The total effective rate in the control group and in the observation group was respectively 57.62% and 94.75%, and the physical function in the observation group was obviously better than that in the control group, and the difference was statistically significant (P<0.05). Conclusion: The zinc sulfate in treatment of patients with diabetic peripheral neuropathy can effectively improve various symptoms of patients, relieve the pains and improve the treatment effect, and it is of clinical application significance.
It can be seen from the above that 3 web translation Apps are pretty good for paper at basic level, where Google and Bing are more loyal to sentence structure and original words order while Baidu seems flexible for its unnecessary use of attribute clause.
3. Academic Paper (intensive)
Let's see some more intensive and professional:
The common conductor is connected to the protective bonding circuit at the point of supply. All contacts, solid state elements etc., which are intended to operate an electromagnetic or other device (for example, a relay, indicator light) are inserted between one side, the switched conductor of the control circuit supply and one terminal of the coil or device. The other terminal of the coil or device (preferably always having the same marking) is connected directly to the common conductor of the control circuit supply without any switching elements (see Figure 3).
From the above test, we found that except Bing Translator, which has made severe mistake on sentence structure, Google and Baidu are extraordinary in translation accuracy and sentence cohesiveness. It should be noted that compound nouns with multiple modifier such as "控制电路电源有开关的导线一边" (switched conductor side of the control circuit power supply) are better translated from Baidu than that from Google.
Texts in literature are poetic, ambiguous, connotative, and irreplaceable in volcabulary. For this reason, it is definitely challenging for translation programs even human beings to convert from original literature text into any foreign language. Let's see the machines' performance:
Shall I compare thee to a summer’s day?
Thou art more lovely and more temperate:
Rough winds do shake the darling buds of May,
And summer’s lease hath all too short a date;
Sometime too hot the eye of heaven shines,
And often is his gold complexion dimm'd;
And every fair from fair sometime declines,
By chance or nature’s changing course untrimm'd;
But thy eternal summer shall not fade,
Nor lose possession of that fair thou ow’st;
Nor shall death brag thou wander’st in his shade,
When in eternal lines to time thou grow’st:
So long as men can breathe or eyes can see,
So long lives this, and this gives life to thee.
Shakespeare's words are just untranslatable. Over the years, people never stopped their efforts to research on this greatest poet and playwright's life and works. We found that when translated into Chinese, the poem lost its expressive power even those meaning of its sentences remains the same.
Artificial Intelligence is amazing. Whether Google or Baidu, they are both doing a great job on machine learning and translation. For normal users who are merely seeking for decent solutions for document translate, a web translator like Baidu translate or Google translate is enough for roughly using. What's more, we have enough reasons to believe that in the future, jobs of translators will be totally replaced by AI.