技术应用 | 外语教师如何使用AI工具来制作英文微课
The following article is from 简言 Author 韩林涛
在这篇文章中我将基于Perplexity、Gamma、ElevenLabs、剪映等工具来介绍外语教师如何使用AI来制作英文微课。
第一步：使用Perplexity撰写内容大纲
发出问题后，Perplexity进一步让我选择感兴趣的领域：
随后，Perplexity开始展示获取到的数据来源，并基于这些来源开始生成答案：
如下：
## Introduction to Computer-aided Translation History
Computer-aided Translation (CAT) has evolved significantly since its inception, becoming an integral part of the translation process. The history of CAT can be traced back to the 1960s, with the development of translation technology marking a significant shift in how translation was approached and executed. The early stages of CAT were characterized by the development of machine translation (MT), which sought to automate the translation process entirely. However, it soon became apparent that fully automated translation was fraught with challenges, leading to the emergence of CAT as a tool to assist human translators rather than replace them[1].
## Key Developments in CAT
The development of CAT has been marked by several key milestones:
- **Early Machine Translation Efforts (1967-2013)**: The initial focus on machine translation highlighted the potential and limitations of automated translation. Pioneering work in this area laid the groundwork for subsequent developments in CAT[1].
- **Translation Memory Systems**: One of the most significant advancements in CAT was the introduction of translation memory (TM) systems. These systems store previously translated segments, which can be reused in new translations, thereby increasing efficiency and consistency[1][8].
- **Integration of Statistical Methods**: The incorporation of statistical machine translation (SMT) represented a major shift in CAT. By utilizing large corpora of bilingual text, SMT systems could produce translations based on statistical models, which improved over time with more data[1].
- **Advent of Neural Machine Translation (NMT)**: More recently, the field has seen the rise of neural machine translation, which uses deep learning algorithms to model the entire translation process as a single, large neural network, leading to translations that are often more fluent and accurate than those produced by SMT[14].
- **Development of Specialized CAT Tools**: Various tools have been developed to address specific aspects of the translation process, such as terminology management, quality control, and project management, further enhancing the capabilities of CAT[7][11][20].
## Challenges in CAT
Despite the advancements, CAT faces several challenges:
- **Language Fluency**: Achieving fluency in second languages using CAT tools remains a challenge, particularly in educational settings where students need to acquire the subtleties of academic language[3][9].
- **Quality and Reliability**: Concerns about the quality and reliability of translations produced by CAT tools persist, especially when dealing with complex languages or specialized terminology[12].
- **Psychological and Physiological Impact**: The introduction of CAT in educational settings has shown that students may face psychological and physiological challenges that affect their learning experience and confidence in translation[4].
- **Integration with Human Skills**: Balancing the capabilities of CAT tools with the nuanced skills of human translators is an ongoing challenge, as is the need for translators to adapt to rapidly evolving technologies[15].
## Applications of CAT
CAT has found applications in various domains:
- **Education**: CAT tools are used to support students learning English as an Additional Language (EAL) and to help them achieve the required standards in English across the curriculum[3].
- **Professional Translation**: CAT is widely used in professional translation, particularly in technical fields where consistency and efficiency are paramount[2][16].
- **Language Service Industry**: The language service industry has embraced CAT for its ability to improve translation efficiency and reduce costs, with applications ranging from software localization to technical documentation[5].
- **Online Learning**: CAT systems are increasingly being applied in online learning environments, facilitating translation tasks and supporting the learning process[18].
In conclusion, the history of CAT is characterized by a continuous interplay between technological advancements and the evolving needs of translators. As CAT tools become more sophisticated, they offer greater support to translators, but also present new challenges that must be addressed. Educators teaching the history of CAT should focus on these key developments, challenges, and applications to provide students with a comprehensive understanding of the field.
Citations:
[1] https://www.semanticscholar.org/paper/746d6ece244ad5288146c18580c8e1651b99649d
[2] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9470352/
[3] https://www.semanticscholar.org/paper/fc9391e86197944db398fb2611773329225b2788
[4] https://www.semanticscholar.org/paper/3538e97c762d933076fb2af30b642ad3d027076f
[5] https://www.semanticscholar.org/paper/34b1a44cea8145b809c11e546f3a8e9c3fedbe2a
[6] https://www.semanticscholar.org/paper/1c24a0eab7d4eb0b780732b7f3be87085ce12aef
[7] https://www.semanticscholar.org/paper/4892d4723058e9313d9c09753b2d04a45774d920
[8] https://www.semanticscholar.org/paper/2d92585962a5cb7f2c1db7196ea3e1a1a77a928b
[9] https://www.semanticscholar.org/paper/5758609c86cee7107c691e3956e480399c1ec579
[10] https://www.semanticscholar.org/paper/e10d0f19761621d7f408f905c119efe673505df5
[11] https://www.semanticscholar.org/paper/63d9f494f23ae7317d8bf1e2885c0b601a277638
[12] https://www.semanticscholar.org/paper/134c446d0004dc8137ed4d4007aef61ab8c40b3d
[13] https://www.semanticscholar.org/paper/b96aa6a35725f8c96367c557be043de3cd8bafca
[14] https://www.semanticscholar.org/paper/bc536fd96e3746df6a9266914b11e69d5319c0da
[15] https://www.semanticscholar.org/paper/6e24d08335213ff2b0aa8c78fc009def5619e513
[16] https://www.semanticscholar.org/paper/9cc27313c5a74bbaf92c17b883415c10223cfb11
[17] https://www.semanticscholar.org/paper/06ff36ce603061d6fad2ff54c7358de57c70e327
[18] https://www.semanticscholar.org/paper/18f27069d2d78620fc31c75a99e7a1144b078a4a
[19] https://www.semanticscholar.org/paper/9aaffc01154b654ac5b8abbe2dc9245120fc7c85
[20] https://www.semanticscholar.org/paper/b0313c0dbc2a22f46f8623b0506b8935f904ff4f
以上内容全部由Perplexity生成。
第二步：在Gamma中生成课件
打开Gamma，将Perplexity生成的文本全部粘贴进入编辑器，选择“Presentation”模式，并点击“Continue”：
Gamma默认选择生成8张课件，我们可以选择课件内容是否需要详细按照文本内容，是否以英语呈现，使用使用AI生成的图片等等。
我们使用默认的选项，并点击“Continue”：
随后选择一个主题，并开始生成：
以下为8张课件的内容：
第三步：使用ElevenLabs生成配音
ElevenLabs可以直接将Perplexity生成的原始文本转换成音频，也可以使用编辑器来自动编辑配音脚本：
如果涉及到PPT中有的内容，但是原始文本没有的，可以让GPT基于PPT生成介绍段落：
最后完整导出所有配音：
这里我使用了自己的声音来作为训练数据，所以英文配音的音色跟我自己的很像。
第四步：将音频和PPT结合
这里我没有找到特别好的自动将音频和PPT结合的方法，虽然可以将音频直接插入到PPT中并导出一个视频，但是很难做到音频播放的声音和某张PPT保持一致，所以我采取的办法就是直接把PPT转成图片，然后插入到视频编辑器中，根据音频的内容来调整图片显示的时长。
将音频和图片导入剪映：
导入音频后，我会先使用剪映自带的语音识别工具生成字幕，方便我定位音频位置：
全部对应好音频和课件后导出：
这个对应的过程比较耗时，得用了20分钟，希望大家能够推荐比较好用的AI工具可以自动实现对齐。
导出后的视频如下：
复盘反思
1）使用AI辅助制作外语微课的成本
在上面的流程中我使用的工具分别有：
Perplexity：这个工具整合了互联网搜索引擎数据和大语言模型，所以在生成答案时可以提供信源，这极大增强了AI生成文本的可信度。这个工具并不对国内开放，且如果要使用高阶功能，需要每月支付20美元。
Gamma：这个工具可以借助大语言模型来将文本拆分成片段，并将片段整合成课件，可以极大减轻教师制作课件的时间成本。但是这个工具依然不对国内开放，且如果要使用高阶功能，需要每月支付20美元。
ElevenLabs：这个工具可以使用很少的个人音频素材训练发声引擎，而且效果奇好。但是这个工具依然不对国内开放，且如果要使用高阶功能，需要每月支付22美元。
剪映：哦，可爱的剪映，是这次实验中唯一不收费的工具，且唯一可以面向国内用户开放的工具。
时间成本：本教程的前三步加在一起的时间大概是15分钟，剪映合成音频和图片所花的时间是20分钟，加在一起一共花费了35分钟。
2）使用AI生成内容的问题
首先，作为计算机辅助翻译领域的教师和研究者，我其实不会按照AI生成的内容来讲解计算机辅助翻译发展的历史，因为我觉得Perplexity生成的内容虽然没有语法错误，但是丢失了太多的细节，可以用来练习听力和口语，但是不能作为专业的授课内容。
其次，AI生成的内容一定要认真检查和审核。比如如果大家仔细听上述视频的8:05左右，会发现AI突然生成了一段奇怪的女声，而并没有使用我自己的声音，所以一下子特别吓人。我把这段保留了下来，让大家感受一下为什么AI生成的内容不能直接使用。
3）未来可能会发生什么
其实站在2024年的4月来谈未来是非常迷幻的，因为过去一年半发生的事情就让人觉得已经生活在未来了。
我可能睡了一觉之后知识和技能都只是略微增长了一点，甚至完全没有增长，还会倒退。
但AI的发展却真的是无时无刻不在飞速推进。所以我并不觉得我今天发现的AI的缺点会一直存在下去，而是会很快更新和升级换代。
所以，我自己会一直跟随AI发展的步伐，看看哪些可用，哪些不可用，并尽量尝试在自己的教学中合理使用，以及引导学生合理使用。
4）AI是否会替代教师？
我觉得随着我对AI的不断深入理解，我和它实际上是共同成长的。但肯定跑不过AI，AI替代我是早晚的事情，所以我一定会在每个工作日的9点-17点之间好好努力工作。
人的生命其实很短暂，AI的发展肯定会延续下去，我更希望我的下一代能够站在AI的肩膀上去更好的探索世界的美好和惊奇。
群内会定期推送语言服务行业最新动态、活动预告、竞赛通知📝等内容~
欢迎你的加入🥰！
往期回顾
2. 行业资讯 | 新增山东省高校！已有19所高校申请翻译博士专业学位
3. 行业动态 | 中国翻译协会医学翻译委员会2024年第一次主任会议在京召开4. 行业资讯 | “新汉学计划”欢迎加入！2024年外语翻译硕士培养项目开始招生5. 行业动态 | “人工智能+”首次写入《政府工作报告》，大模型如何助力产业创新？
行业洞见
行业洞见 | 《翻译技术研究》作者王华树教授 行业洞见 | 为腾飞助力：AI 如何为翻译引擎增加动力 行业观察 | 人工智能教父谈人工智能的影响和潜力 行业观察 | 奥特曼专访：关于 GPT-5、Sora、Ilya、Q*、AGI、外星人等等一切
行业技术
行业技术 | 带你了解大语言模型的前世今生 技术应用｜Mac和Win如何协作制作双语平行语料库？ 技术应用 | 翻译实践中术语库的应用 技术应用 | 术语库创建与应用 技术应用 | 7个技巧让你成为谷歌Gemini专家
精品课程资源干货
技术科普 | 揭开手写文本识别“最先进”的秘密 资源干货 | 英语外刊获取途径、网站推荐 ChatGPT | AI外语写作助手，助力高效写作 双语干货 | 谢锋大使向中美媒体发表讲话 资源宝库｜译者用“典”，多多益善——精选31款在线词典
招聘就业1. 译者招募丨英译中，经济学（有署名）2. 招聘快报 | CNBC财经电视台上海分公司招聘发行部助理实习生