机器翻译系统要正确翻译，是以正确理解为前提的。今天在阅读文献时，用机器翻译来辅助阅读，发现机器翻译翻译错了。仔细分析了一下，发现错误的原因是机器翻译未能理解一个并列结。构。

Sentences from CNN

1. Despite being the first UK premier in history found to have broken the law in office, and people working in Downing Street receiving hundreds of police fines for breaching Covid regulations that Johnson himself made, the end of this police investigation significantly reduces the risk of another smoking gun the PM's enemies can use to get rid of him.

尽管是英国历史上第一个被发现在任职期间违法的总理，而且在唐宁街工作的人因违反约翰逊本人制定的科维德条例而收到数百张警方罚单，但这次警方调查的结束大大降低了总理的敌人可以用来摆脱他的另一把烟枪的风险。(DeepL)【句子的主干（main stem）为划线部分。导致该句变得长而复杂的不仅仅是两个加粗变斜体的关系从句，还有despite后面两个并列的nominal phrases。】

2. London (CNN)Boris Johnson has been able to end what might have otherwise been a troubling week with an unexpected sigh of relief.

伦敦（CNN）鲍里斯-约翰逊（B

      在看学生的翻译报告，发现一个带定语从句的句子翻译得比较别扭。

       （Trust within this model, therefore, relies largely on accountability, efficiency, and adherence to contract.） Law, within this paradigm, removes any of the true need to cement business dealings into relationships that outlast adversity or weather business setbacks.

      我们在研究翻译和英语以及英语教学中，常常需要一个规模较大的语料作为参考语料库（reference corpus）。常用的参考语料库有BNC，BROWN等等。Mark Davis所创建的系列语料库中有一个当代美国英语语料库（COCA），其规模有数亿词，涉及不同的语类，能够很好地代表当代美国英语。因此，在英语研究和教学中，选择COCA作为参考语料库是一个比较合理的选择。

       但是，COCA是一个在线语料库，怎么获得其词表呢？平时没怎么注意，前几天有一位网友问到如何获得COCA词表这一问题，我才去寻找。现将其获取地址公布如下：

地之一：原网址下载：https://www.wordfrequency.info/files/entriesWithoutCollocates.txt

百度网盘分享：链接：https://pan.baidu.com/s/1Xga49Ayp-JuDZfPgyibUDQ?pwd=lana
提取码：lana

      

      机器翻译和人类翻译表面上类似，从产品（product）层面来说，最终大家都将翻译的结果摆在面前。机器翻译的表达是显而易见的，毕竟有一个结果在这儿。虽然和人类翻译相比存在差别，但是，其结果应该是表达的体现。理解又怎么去认识它呢？按道理，机器翻译也存在理解阶段。说它有理解这个阶段我们是从人类翻译这个角度去理解机器翻译的推理，或者说是一个假设。如果说人类翻译过程时一个黑箱，机器翻译的黑箱可能比人类翻译的黑箱容易理解一点，因为，机器翻译系统毕竟是人类构建的系统，如果其中有理解这个过程和实施方式，系统工程师应该有明确的方案。只不过机器翻译的理解过程可能和人类理解过程存在较大的差异。

       郭子凯等（2022:193）在评价机器翻译和人工翻译差异时，说到：“事实上，机译译文虽然在译文表达层面尚有比较大的改进空间，但在对句型的理解准确性上令人称道。”似乎是机器翻译理解没有问题，我们看到的问题只不过是表达上的问题。从表面上看，可能是机译系统表达的问题，但是，这样的问题是不是可以说，是机器翻译系统没有理解，所以才导致我们看到翻译结果上的问题

     正在上King's English and Pub Talk。其中有一个词汇学习任务，练习 要求翻译一个汉语句子，并且用上indulge。在交上来的练习中大多数学生使用的indulge in+wine/it，这也是课文中的用法。但是，有极少数的学生这样译 

       I have being prompted by questions from my colleagues that the symbole of # in the wordlist generated by Wordsmith Tools. I have tried many times with an attempt to find out the cause of this error, but failed until this afternoon when another guy from bilbili (B站) threw me the same problem. Again I played with my WST, though I was not sure if I could find any solution. I was even not sure if I cound solve this problem. From what I have done before and the experience of the guy who sent me the messege today, I felt quite sure the problem is closely associated with the demonstration of number. I still remember the guy told me there is no # in the wordlist after the numbers in the corpus have been deleted. This message from the guy confirmed that this might be the cause, though I would not do the same just to hide the # in the wordlist.

      I kept playing with my WST, especially the settings of languages.

       US and its allies are posing sanctions against Russia for its military operation in Ukraine. Thefull range of sanctions covers government officials and civilians, schools and private companies, including cats, dogs, even trees in Russia.  To the surprise of many westerners, Russia launched its anti-sanction measures in response.

       Many people are worrying about the American astronaut who is scheduled to come back to earh. Is the Russian space agency going to leave the American astronaut in the space station as a anti-sanction to US? However, what is beyond the expectation is that the Russian promised to take the American astronaut back as usual.

      What is more interesting is that in the piece of news, the same 'astronaut' referred to in different words respectively, the American astronaut, the Russian cosmonaut.