From ab53ed6bd06162b06a0b4780343d995b6956786a Mon Sep 17 00:00:00 2001 From: github-actions <41898282+github-actions[bot]@users.noreply.github.com> Date: Fri, 19 Jul 2024 22:10:39 +0000 Subject: [PATCH] [U] Update channel content --- exports/hykilp/atom.xml | 2 +- exports/hykilp/index.html | 2 +- exports/hykilp/posts.json | 10 +++++----- exports/hykilp/rss.xml | 2 +- 4 files changed, 8 insertions(+), 8 deletions(-) diff --git a/exports/hykilp/atom.xml b/exports/hykilp/atom.xml index 491cdcf30..74121fe76 100644 --- a/exports/hykilp/atom.xml +++ b/exports/hykilp/atom.xml @@ -2,7 +2,7 @@ https://aza.moe/life 小桂桂的回忆录 📒 - 2024-07-19T20:11:03.523855+00:00 + 2024-07-19T22:10:37.300959+00:00 python-feedgen https://aza.moe/meru_256px.png diff --git a/exports/hykilp/index.html b/exports/hykilp/index.html index 001a37ae1..0a039c2ba 100644 --- a/exports/hykilp/index.html +++ b/exports/hykilp/index.html @@ -27,7 +27,7 @@ diff --git a/exports/hykilp/posts.json b/exports/hykilp/posts.json index 2c8a70df6..37a8446f2 100644 --- a/exports/hykilp/posts.json +++ b/exports/hykilp/posts.json @@ -56874,7 +56874,7 @@ "id": 4172, "date": "2024-07-19T10:57:17", "text": "跟我学单词:\nCrowd,人群\nStrike,罢工\nCrowdStrike,集体罢工\n\n例句:\nJuly 19th, we CrowdStrike. 7月19日,我们集体罢工了", - "views": 40, + "views": 41, "forwards": 1, "forwarded_from": { "name": "IloveMATH.Get('Daily').Except('Ingress').EmitToChannel()" @@ -56884,7 +56884,7 @@ "id": 4173, "date": "2024-07-19T10:58:45", "text": "感觉会有很多一般路过吃瓜群众觉得这次的蓝屏事故是微软的锅...", - "views": 48, + "views": 49, "forwards": 0, "reply": { "id": 4172, @@ -56895,7 +56895,7 @@ "id": 4174, "date": "2024-07-19T14:30:25", "text": "今天在研究中日文翻译,去🤗上找了一圈发现并没有人训练中文翻日文的模型,即使语言很全的比如 Helsinki-NLP opus 系列也大部分都是转英语的模型。\n\n然后就在想,要做一个语言覆盖全面的翻译模型需要的资源好多啊,模型数量和语言数量是 n² 关系,如果要支持 33 种语言互相翻译就需要训练 1089 个模型... 那现有的翻译工具是怎么做的呢?\n\n然后发现谷歌翻译其实是用英语做了中间语言,先把所有语言翻译到英语再翻译到目标语言... 这样就只需要 2n 个模型了,但是会有很大问题,英语有歧义的东西就会翻错,比如「字符串」 > \"String\" > 「弦」\n\n那 DeepL 是怎么做的呢?简单测试一下发现是没有英语歧义问题的,字符串不会翻译成弦,弦也不会翻译成字符串... 但是仔细测试一下发现 DeepL 其实也是用英语中转的,因为中英翻译爆炸的时候中日会把英语的爆炸结果翻出来。那它是怎么做到消歧义的呢?好奇妙", - "views": 42, + "views": 43, "forwards": 0, "media_group_id": 13771195402520333, "images": [ @@ -56935,7 +56935,7 @@ "id": 4177, "date": "2024-07-19T14:42:19", "text": "没事了,DeepL 英语歧义也会翻错", - "views": 37, + "views": 38, "forwards": 0, "reply": { "id": 4174, @@ -56959,7 +56959,7 @@ "id": 4178, "date": "2024-07-19T15:36:21", "text": "测试了半天百度翻译,百度翻译确实没有英语歧义的问题\n\n...因为百度的中转语言是中文,当然没有英语歧义的问题,只有中文歧义的问题 🌚\n\n\"Beat a car\" > 「打车」 > 「タクシーを拾う」 (\"Take a taxi\")\n\n不过感觉中文确实比英文更难出现歧义,而且因为亚洲语言离中文更近,感觉百度翻译翻亚洲语言的准确率肯定比英语中转的 Google / DeepL 高", - "views": 30, + "views": 31, "forwards": 0, "reply": { "id": 4174, diff --git a/exports/hykilp/rss.xml b/exports/hykilp/rss.xml index df8a317f8..76f6a0e70 100644 --- a/exports/hykilp/rss.xml +++ b/exports/hykilp/rss.xml @@ -12,7 +12,7 @@ https://aza.moe/life zh-cn - Fri, 19 Jul 2024 20:11:03 +0000 + Fri, 19 Jul 2024 22:10:37 +0000 小桂桂的回忆录 📒 #4178 https://aza.moe/life?post=4178