每天给你送来NLP技术干货！
来自：看个通俗理解吧
有趣大会 · ACL2022 (Findings篇)
Interesting · ACL2022 (Findings)
什么是有趣“大会”？它并不是一个和官方相关，真正的大会。它会是以趣味为导向的一系列文章，可以用放松、不太正式的心态去阅读它。
What is the series of "Interesting"？It will be a series of entertainment-oriented articles. These articles can be read with a relaxed, less formal mindset.
有趣大会都会有哪些内容，可以给我带来什么？ 会挑选会议中一些有奇奇妙妙的想法或者任务的工作，然后进行非常简短的介绍。定位在于奇妙的角度或者有趣的任务，所以不会出现对模型方法的深度解读。但是，希望它能够为你的工作和学习中带来一些放松，更希望能帮你打开思路，用最小的精力和时间开销去看到更多形形色色、并且是在领域内的研究。后续可能会出现有趣大会之*ACL，EMNLP，COLING等。
What will the series of articles cover and what benefits does it bring me？ These articles will present a selection of studies from the top conferences with interesting ideas or tasks, followed by very short explanations. There will not be in-depth explanations of models and methods. Hopefully, this series will help you to relax from your heavy work and studies. Moreover, we also hope the ideas and tasks can give you more inspiration, broaden your mind, and help you to understand the general ideas of interesting studies in the field with minimal effort. In addition to "Interesting · ACL2022", we may publish more articles such as "Interesting · *ACL, EMNLP, COLING" etc.
声明：
内容仅代表个人观点
个人难免会出现遗漏，欢迎小伙伴们私信提出意见或者补充
Statement：
The content represents personal views only
We may have missed some papers, thus greatly appreciate your valuable comments
【有趣大会 · ACL2022 （I·ACL2022）】
Findings篇 (Findings)(←)
短论文篇 (Short Papers)
长论文篇 (Long Papers)
Findings
1. 音乐博主的潜在流量密码（Secret to Potential Popular Music-Related We-Media）
Automatic Song Translation for Tonal Languages
https://aclanthology.org/2022.findings-acl.60/
论文配图 (Figure from the paper)：
你还记得那首上头、洗脑的日文直译版《好汉歌》嘛？可以想象的出，作者为了填词应该下了不少的功夫，尝试了无数次，脱了大把的头发（不是，开玩笑可别当真呀）。而这篇论文，为音调语言开发了自动歌曲翻译器，解决了很重要的一个问题：在能传达原始含义的同时，还可以将词的音调与歌曲的旋律相一致。这无疑将会大大的提高音乐创作者在填词方面的效率。
There is a song coming from China and there is a person who translated the Chinese lyrics directly into Japanese（https://www.bilibili.com/video/BV1UX4y1P76a/）. The sticky cover song has caught the attention of many people because it sounds very funny. As you can imagine, the author must have put a lot of effort into writing the lyrics, revised numerous times and lost a lot of hair (no, don't take it seriously). This paper develops an automatic song translator for tonal language and solves a very important problem: its output can convey the original meaning while at the same time aligning the words' tones with the melody of the song. This will undoubtedly greatly increase the efficiency of music creators in terms of editing lyrics.
“什么是音调语言（what is a tonal language）？：音调语言的定义是，不同的词用不同的音调转折会表达不同的意思。例如，一个词可以用四种不同的声调说出来，而每一种声调都会改变这个词的含义。A tonal language is defined as a language where different words with different tonal inflections will convey different meanings. For example, a single word could be said with four different tones, and each of those tones will change the meaning of the word. （Source: https://akorbi.com/akorbi-explains-tonal-languages/)）
”
2. 时空失控“恋人” （Dislocation in Time and Space）
Good Night at 4 pm?! Time Expressions in Different Cultures
https://aclanthology.org/2022.findings-acl.224/
论文配图 (Figure from the paper)：
在人与人沟通之中，对对方的话产生错误的解读，会导致大大小小的不好的结果。说同一种语言的人，在一起交流都会产生误会 ，更不要说跨国家的情况了。这篇论文从很小的一个点出发，开发了一种可以把时间描述转换成对应的文化中对应的时间区间。如果把这样的功能用在更多的不同的语言，会产生什么样的效果呢？
In human communication, misinterpretation of the other person's words may lead to very bad results. Even though two people are of the same cultural background, they may misunderstand each other's words, not to mention two people with different cultural backgrounds. This paper focuses on a very clever point and develops a method to convert time expressions into corresponding time intervals in the culture. What would be the effect of applying such a tool in more languages?
“论文摘要中对这个任务是什么举了一个例子（example of what is the task taken from the paper）：mapping from expressions such as “morning” in English or “Manhã” in Portuguese to specific hours in the day
”
3. 可以代替你爬楼的机器人 （Solution Hunter Bot）
Learning to Describe Solutions for Bug Reports Based on Developer Discussions
https://aclanthology.org/2022.findings-acl.231/
论文配图 (Figure from the paper)：
"求求了，翻翻上面的聊天记录吧"，"刚才刚说了，真的说累了。办法就在上面，自己看吧。"，"我就是这样解决的，太好了！（底下回复：你用的哪个方法啊？可以说说吗？）"... 这样的讨论每天可能在很多地方都有发生。往往一个问题的抛出，会引出来很多的讨论。再加上一些小伙伴思维比较发散，说些有的没的，有价值的信息就会被埋没。有了这个机器人就不用怕了，它会综合大家说出的话，然后自动总结出来解决方案是什么。省去了自己爬楼的时间。要是把这样一个机器人放到聊天答疑群里面，它会总结出来什么呢？
"Please, go through the chat history above", "I just said that. I'm really tired of saying it again.", "I solved it as suggested, great! (Reply: what solution did you use? Can you tell me more details?") ... This kind of discussion can happen in many places every day. Usually, a question is thrown out, which will lead to more problems for discussion. Even worse, some people often post messages that are not directly relevant to the problem. That can result in useful and valuable information being buried. However, the method proposed by this paper can go through what everyone said and then automatically summarise what the solution is. Therefore, it saves you a great deal of time in finding solutions by yourself. If we integrate a bot like this into question&answering chat groups, how will it behave?
下一篇 (Next)
下一篇会进入有趣大会的短论文篇章~
We will move on to the Interesting · ACL2022 (Short Papers) session~
📝论文解读投稿，让你的文章被更多不同背景、不同方向的人看到，不被石沉大海，或许还能增加不少引用的呦~ 投稿加下面微信备注“投稿”即可。
最近文章
COLING'22 | SelfMix：针对带噪数据集的半监督学习方法
ACMMM 2022 | 首个针对跨语言跨模态检索的噪声鲁棒研究工作
ACM MM 2022 Oral | PRVR: 新的文本到视频跨模态检索子任务
投稿或交流学习，备注：昵称-学校（公司）-方向，进入DL&NLP交流群。
方向有很多：机器学习、深度学习，python，情感分析、意见挖掘、句法分析、机器翻译、人机对话、知识图谱、语音识别等。
记得备注~
210
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
