⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

💡 Is anyone willing to help mediate this dispute?

Please feel free to write to me at angeci (at) ltgc.cc and discuss this issue together!

Of course, it would be best if this were simply a misunderstanding and someone could help mediate, but the current evidence suggests otherwise. I will not retract my statement for now, as his viewpoint (or rather, his inappropriate analogy) truly offends me.

Alex Hsu mentioned that he was raised in mainland China, so it’s no surprise that he supports the Chinese Communist Party.

I found that one of Alex Hsu’s article “How Taiwan and China rename English movies (and why it’s a mess)” has the implication of supporting the suppress of local languages with Mandarin Chinese, and I wrote a response article “Unifying Names” about it. I initially thought he had simply overlooked something. I added a supplementary explanation in the form of a comment when his website still had a message board, but a few days later he deleted my previous explanation along with the entire message board. I tried writing to him to discuss this issue, but my dissatisfaction ultimately went unanswered. This confirms that Alex is indeed a member of the CCP, a proponent of Mandarin hegemony, and a proponent of Mandarin chauvinism, and he is a fellow traveller of the CCP. Next, I will update my previous response article on my attitude towards Alex, and forcibly convert all direct quotes about Alex on the website to simplified Chinese (embedding simplified Chinese within traditional Chinese can be seen as a unique rhetorical device).

I know that there are people in the BlogBlog Club who also support the preservation of Taiwanese and other local languages, and some even submit their posts using only romanized Taiwanese¹. You should also unite, point your guns at one enemy, and regard Alex Hsu as your common enemy!

Basic command

yt-dlp [param1] [param2] [param3] "URL 1" "URL 2" "URL 3"...

In addition to providing the URL for a single video, the URL of a playlist can also be used for batch downloading. This technique “may” work for platforms other than YouTube, but I haven’t tried it myself.

The --output parameter can be used to specify the output filename. If you don’t specify it, it will automatically fetch the video title and ID for you. I almost never specify the output filename by myself.

yt-dlp --output "%(title)s.%(ext)s" <URL>

This will remove the video ID from the output file name, leaving only the video title.

The -P parameter can be used to specify the output path. I currently rarely download in batches so I rarely change this setting.

Search for available video formats

yt-dlp <URL> -F

If you use both -F and -f, followed by a format filter, it will list the available formats for the current video based on the filter you specified (or all available formats if no filter is specified), but it will not download the video.

Custom codec priority

yt-dlp thinks that the video codec AV1 is better than VP9 by default, but there’s a problem for AV1, which has a poor software decoding performance, resulting in a poor playback experience on some devices. Furthermore, I rarely need to access ultra-high resolution videos, so I’d like to lower the priority of AV1 encoding (at least to below H.265?):¹

yt-dlp <URL> -f (271/248)+251

This is a code snippet I hastily wrote before. However, this command only works if the target video provides one of the above-mentioned formats; otherwise, it will directly report an error.

In practice, if the video source is YouTube, the default audio format fetched is mostly the opus format numbered 251. Therefore, the audio format part can perhaps be directly replaced with ba or bestaudio.

yt-dlp <URL> -S "vcodec:vp9"
yt-dlp <URL> -f "bv*[vcodec!=av01]+ba/b"

Custom download list

If the videos you want to download in batches are not already in the same playlist, you can also write the desired download list as a text file for yt-dlp to read:

yt-dlp --batch-file urls.txt

Local preset options

• Linux/macOS path: ~/.config/yt-dlp/config/yt-dlp.conf
• Windows path: %APPDATA%\yt-dlp\config\yt-dlp.conf; Or simply put config.txt under the same directory of yt-dlp.exe.

Other options

• --skip-download: Skips the download process for videos. Usually used for fetching thumbnails or metadata.
• --add-metadata: Embeds the video’s metadata into the synopsis field of the video file, or the performer information in the audio file. --write-description outputs the description field as a text file.
• --embed-thumbnail: Attempts to embed the video thumbnail into the video file (provided the container format supports it). --write-thumbnail directly downloads the video thumbnail.
• --embed-subs and --write-subs options are used to download external subtitles. The --sub-lang option specifies the subtitle language. By default, external subtitles are not downloaded unless these two parameters are explicitly specified.
• --list-subs: Lists all available subtitles.
• --keep-video: Keep the orginal files after automatically merging the video and audio parts.
• --split-chapters: Split the output video into chapters automatically based on its chapter information.
• -o -: To output downloaded videos to stdout, generally used together with piping.
• --cookies-from-browser <browser>: If the downloaded target requires login for access, then this parameter is needed to feed the necessary cookies to yt-dlp. The name of a browser should follow this parameter, or you can also export the cookies manually.
• --write-comments: Downloads video comments.

Other random thoughts

• Many years ago, I tried to fetch multi-part videos from Bilibili, but it could only fetch P1 from it. I haven’t tried this again at a later time.
• Downloading Twitch live stream archives is very slow in practice, contacting the streamer directly for a download link might be way faster.

I recently watched this video, there’s a comment on that video that says:

问题是，我见过的路痴，就是跟着地图，导航走也会迷路😂

比如我老婆，她看导航，那么大一张地图她只能理解一个数字：距离目标地点多少米😂。

她跟导航的方式是：先随机找个方向走，然后看目标距离是增加还是减少。如果发现增加了就随机再换个方向。如果发现减少了就一直走下去😂。

——某 YouTube 用戶的評論

This kind of navigation technique has a special term in computer science, which is called “gradient descent”.

🚨 Caution

This article about discussion on the Chinese language itself and is not translated in to English yet (and probably never), please refer to the Chinese version instead.

This page is still work in progress, stay focus for continuous upcoming updates on this page!

聲母

早期中古漢語

• 早期中音漢語大概有 38 個聲母音位。在古代，聲母在寫作韻文時的地位不如韻母重要，導致在韻圖出現之前的文獻往往對聲母不太重視。《切韻》時代的聲母只能靠繫聯法或外語對音等手段推出。不同文獻列出的聲母數量有所出入，從 35 個到 51 個聲母都有。
• 聲母按發音部位傳統上分為「唇舌齒牙喉」五種。有些發音部位還可以再細分形成舌頭、舌上、半舌、齒頭、正齒、半齒等發音部位。
• 唇音、牙音、喉音合稱為鈍音聲母；舌音和齒音合稱為銳音聲母。
• 聲母按發音方式傳統上分為全清、次清、全濁、次濁。這些發音方式的差異在後世通常反映為聲調的分化，保留古代濁塞音和濁塞擦音聲母的現代漢語方言並不多（但不少方言有後起的濁塞音和濁塞擦音聲母，通常從鼻音變來）。
• 牙音聲母「見谿疑」（可能也包括曉母）在某些時期或地域接非三等韻母時，實際發音有可能是小舌音而不是硬腭音，但這兩組發音並不構成音位對立。
• 以下表格為了排列方便，「以來日」的位置被挪了一下。這幾個聲母傳統上的發音部位歸類分別為喉音、半舌音、半齒音。
• 有些材料列出的聲母數量可能有所差異，例如匣母與云母互補、沒多少字的俟母和船母應否獨立等等。

後期中古漢語

• 在韻圖時代，中古漢語的聲母系統發生了變化，形成了廣為流傳的「三十六字母」體系。
• 輕唇音出現，分化條件我寫在韻母的部份裏。/f/ 和 /fʰ/ 的對立在語音學角度來看很不穩定，非敷分立有可能只是形式上的仿古而已。
• 兩組正齒音合併（或者說由於「莊三化二」徹底完成的緣故，莊組與章組變成了互補分佈不再對立），但常母和船母在這個時代應該已經相混。在韻圖中，通常常母歸入禪母，而船母歸入牀母。

韻母

韻母結構

• 一個中古漢語韻母可以分析成介音、韻腹、韻尾三個部份。
• 介音傳統上又被分成「等」和「呼」兩個部份。
  • 「呼」有「開口」和「合口」兩種狀態，大致上可以對應於 /w/ 介音的有無。至於「等」的詳細敘述見下。
  • 有趣的是，同一個韻母的開合口變體，只有主元音靠後的一等和三等 C 類，開合口變體才可能有不同的名字（但也有共用一個名字的），其餘主元音靠前的二等、三等 AB 類和四等，開合口變體都是一律共用一個名字的。
• 古代的音韻學文獻很少分析「中古漢語到底有多少個韻腹」這回事，等到「十六攝」的概念出現的時候，整個音系其實已經出現簡化了。在不考慮開合口對立的情況下，中古漢語在大多數情況下擁有五組相互對立的主元音，可以分析成典型的「aeiou」體系。
• 中古漢語的韻尾有八種，連同空韻尾一共有九種狀態。這九種不同的韻尾狀態傳統上分成三類：
  • 陰聲韻：空韻尾、半元音 /i/ 和 /u/
  • 陽聲韻：鼻音 /m/、/n/ 和 /ŋ/
  • 入聲韻：塞音 /p/、/t/ 和 /k/
• 古代韻書有「韻目」的概念，本質上就是把各種「韻母 + 聲調」的組合編成一個個韻目而已。

等

在韻圖時代，各種韻母按照介音劃分成四個「等」。此外也有根據反切材料推定、適用於早期中古漢語的「反切等」。

從切韻時期的擬音來看，三等帶有一個類似於 /i/ 的介音（其實切韻時期的三等介音還可以再細分），而非三等則按照主元音的前後分出一等（後元音）、二等（央元音，早期可能還有捲舌成份）、四等（前元音）。

韻圖等和反切等的一個主要區別是，一部份反切三等被重新歸入韻圖二等或四等：精組反切三等被歸入韻圖四等、「莊三化二」徹底完成、3C 與 3B 合流形成韻圖三等、四等生出 -i̯- 介音與 3A 合流，形成韻圖四等。韻圖等屬於二等或四等的音節、但在反切等屬於三等的音節，稱為「假二等」和「假四等」（儘管重紐四等通常分類成獨立的一類而不視為假四等的一種）。

早期中古漢語的三等韻母非常多，而且部份韻母出現重紐或者其他的一些複雜現象，故可以再細分成幾類。三等 AB 類主元音為前元音，且在唇音聲母後保留重唇音；三等 C 類主元音為央後元音，且在唇音聲母後變為輕唇音。

• 一等：東冬模泰灰咍魂痕寒桓豪歌戈唐登侯覃談
• 二等：江佳皆夬刪山肴麻庚耕咸銜
• 三等 A 類：「支脂祭真仙宵侵鹽」重紐四等、麻三、清韻、幽韻牙音
• 三等 B 類：「支脂祭真仙宵侵鹽」重紐三等、庚三、幽韻唇音
• 三等 C 類：微廢文欣元嚴凡、東鍾之魚虞戈陽尤
• 四等：齊先蕭青添

韻母的等也和能配什麼聲母有關係，可以說上古的介音共同影響了聲母和韻母的發音。

攝

61 組有單獨名字的韻母，按照韻尾和近似的主元音分類成 16 個「攝」。在廣韻中十六攝的次序是似乎沒啥邏輯的「通江止遇蟹臻山效果假宕梗曾流深咸」。

在韻圖中，「江蟹臻山效假梗咸」被歸類為「外轉」，剩下的「通止遇果宕曾流深」被歸類為「內轉」。只有「外轉」的攝擁有反切二等和反切四等韻。

以下的表格列出了各個攝在韻圖期的擬音。

重韻

同一個攝之內，同一個等的多個韻母稱為「重韻」。這些重韻在後世通常會完全合併以致難以從現代讀音區分。

• 一等重韻三組：東冬、灰咍—泰、覃談
• 二等重韻四組：佳皆夬、刪山、庚二耕、咸銜
• 三等重韻九組：東三鍾、支脂之微、魚虞、祭廢、真諄—文欣、元仙、庚三清、尤幽、鹽—嚴凡

以元音結尾的韻母——果假遇攝

果攝包含「歌哿箇」和「戈果過」兩個韻系，涵蓋了一等和三等各自的開合口韻母，合共四個韻母；假攝僅包含「麻馬禡」一個韻系，但有二等開口、二等合口、三等開口合共三個韻母；遇攝包含「魚語御」、「虞麌遇」、「模姥暮」三個韻系共三個韻母。

之韻系和魚韻系均不配唇音聲母。一些韻圖把魚韻系歸為開口、虞韻系歸為合口，從上古音到中古音的演變規律來看也有一定道理。從上古到中古的演變規律來看，模韻系是開合口中立的，不過在南北朝後期的實際用韻情況是虞模同用、魚大致獨用，顯示模韻的主元音比較接近虞韻而非魚韻。

麻韻三等只能配精組、章組聲母或以母、日母。戈韻系三等字數不多，且只能配牙音聲母。當中包含「迦伽」這些中古時期創造出來專門用於音譯外來語的字，顯示中古時期的音譯習慣認為 /kɨ̯ɑ/ 比 /qɑ/ 更接近梵語 /kɐ/ 的發音（其實不只是戈韻系如此，其他韻母也有這樣的傾向）。

「支脂之微」四個韻系在中古後期合併為單元音 /i/，歸類為止攝；齊韻系和佳韻系屬於蟹攝。

以 /i/ 結尾的韻母——蟹止攝

蟹攝包含「齊薺霽」、「祭」、「泰」、「佳蟹卦」、「皆駭怪」、「夬」、「灰賄隊」、「咍海代」、「廢」合共九個韻系，其中灰咍為開合互補，其餘的韻系都有開合口變體，祭韻另有重紐，合共 18 個韻母。

止攝包含「支紙寘」、「脂旨至」、「之止志」、「微尾未」四個韻系，其中「支脂微」三個韻系各自有開合口之分、「支脂」又有分重紐，實際一共包含 11 個韻母。

以 /i/ 結尾的韻母，差不多把所有「介音 + 主元音」的組合都填滿了。不過由於後世演變的差異，上表列出的韻母和攝的具體範圍有點不同。具體來說，佳韻系本身歸蟹攝，但從後世演變來看，佳韻系實際上被一分為二，一部份後世演變同蟹攝二等，一部份則併入了假攝二等（具體兩類分別包含哪些字存在方言差異）。

「祭泰夬廢」四個韻目只有去聲字而缺少其他聲調，稱為「次入韻」。這些次入韻的主元音高度皆較低，形成機理可以如下解釋：與上古祭部平行的歌部發音變掉了，使得祭部去聲字在中古缺少了對應的平上聲字；但是和上古至隊部平行的脂微部卻沒有類似的變化，於是至隊部的去聲字和脂微部在中古音的對應是一樣的，平上去三聲俱全。

在南朝宋之前，脂之是不能通押的，直到南朝齊才混同。支韻約於唐代才併入脂之。微韻的併入最晚，直到平水韻都是獨用的。

蟹攝方面，平水韻的劃分方式是中前元音的齊祭一組、低後元音的泰獨用、三個二等韻佳皆夬一組、中後元音的灰咍廢一組。

以 /u/ 結尾的韻母——效流攝

效攝包含「蕭篠嘯」、「宵小笑」、「肴巧效」、「豪晧号」四個韻系，其中宵韻系有分重紐，合共五個韻母；流攝包含「尤有宥」、「侯厚候」、「幽黝幼」三個韻系共 3-4 個韻母。

以 /u/ 結尾的韻母均不分開合口。

流攝三個韻系在南北朝後期皆可互相通押，在平水韻裏被一起打包進了同一組韻系。效攝內部的通押性則比較低，在平水韻中是蕭宵同用、肴獨用、豪獨用。

以 /m/ 結尾的韻母——咸深攝

深攝僅包含「侵寑沁緝」一個韻系，有兩個重紐變體，陽入相配合共四個韻母；咸攝包含「覃感勘合」、「談敢闞盍」、「鹽琰豔葉」、「添忝㮇怗」、「咸豏陷洽」、「銜檻鑑狎」、「嚴儼釅業」、「凡梵范乏」八個韻系，其中鹽韻系有重紐，嚴韻系與凡韻系互補，共 16-18 個韻母。

以 /m/ 或 /p/ 結尾的韻母幾乎均不分開合口，除了嚴韻系在唇音聲母後面會變成互補的合口凡韻系之外。

平水韻的劃分方式是，高元音的侵獨用、中後元音的覃談一組、中前元音的鹽添一組、央元音的咸銜與剩下的嚴凡為一組。事實上覃和談可能在南北朝後期就已經混了。

以 /n/ 結尾的韻母——山臻攝

山攝包含「元阮願月」、「寒旱翰曷」、「桓緩換末」、「刪潸諫黠」、「山產襇鎋」、「先銑霰屑」、「仙獮線薛」七個韻系。與蟹攝類似，其中寒桓為開合互補，其餘的韻系都有開合口變體，仙韻另有重紐，加上陽入相配，總共有 28 個不同的韻母；臻攝包含「真軫震質」、「諄準稕術」、「臻𧤛櫬櫛」、「文吻問物」、「欣隱焮迄」、「魂混慁沒」、「痕很恨麧」七個韻系，其中真韻有重紐，文欣、魂痕各自為開合互補；但真諄臻三個韻系是互補的，可以歸為 10 組韻母，合共 16-18 個韻母。

以 /n/ 或 /t/ 結尾的山臻攝大概是把所有「介音 + 主元音」的組合填得最滿的一組韻母。

廣韻的目次中，鎋韻配山韻、黠韻配刪韻，可能是廣韻的訛誤。有些證據可以顯示這兩組關係實際上應該倒過來。

元韻系比較特別，它在中古早期更傾向與魂痕韻通押，所以廣韻的目次排列和平水韻都把元和魂痕排在一起，後世才分出來與仙韻合流。

臻攝的「真諄臻」三個韻系其實是互補的，它們的聲韻組合關係見下表：

平水韻的劃分方式是，高前元音的真諄臻一組、高後元音的文欣一組、中後元音的元魂痕一組、低後元音的寒桓一組、央元音的山刪一組、中前元音的先仙一組。

以 /ŋ/ 結尾的韻母——宕梗曾通江攝

宕攝包含「陽養漾藥」、「唐蕩宕鐸」兩個韻系，這兩個韻系各自是洪細互補，又有開合口之分，加上陽入相配，共有八個韻母。

梗攝包含「庚梗映陌」、「耕耿諍麥」、「清靜勁昔」、「青迥徑錫」四個韻系，四個韻系都各自有開合口，庚韻更區分二等和三等，再加上陽入相配，共 20 個韻母；曾攝包含「蒸拯證職」、「登等嶝德」兩個韻系，各自有開合口。其中三等合口舒聲是缺失的，僅有入聲，共七個韻母。

通攝包含「東董送屋」、「冬湩宋沃」、「鍾腫用燭」三個韻系，其中東韻系有一等和三等兩種變體，冬鍾洪細互補，加上陽入相配，總共有八個韻母。江攝僅包含「江講絳覺」一個韻系，陽入相配合共兩個韻母。

以 /ŋ/ 或 /k/ 結尾的韻母排列得不甚整齊，同屬二等的江韻和耕韻在位置上衝突了。平水韻把這些韻母劃成了八組韻目，超出其他韻尾的「五元音格局」範圍，也許意味着這組韻目實際上應該再分為兩組不同的韻尾。

宕攝、梗攝和曾攝韻母的開合口，只有牙喉音聲母後面存在對立，其他聲母後面一律只能配開口的韻母；而通攝和江攝的韻母就都沒分開合口。

平水韻的劃分方式是，東獨用、冬鍾同用、江獨用、陽唐同用、庚耕清同用、青獨用、蒸登同用。不過中古早期，蒸韻和登韻也是不能通押的。

配列限制

• 唇音聲母一律不分開合口。
• 幫組、見組、影曉母可以配一至四等韻母，但群母只能配三等。
• 端組配一四等，知組配二三等。
• 精組可以配一、三、四等，但邪母只能配三等。
• 莊組只可配二、三等（但俟母只能配三等）；章組、日母、云母、以母均只可配三等。
• 匣母只能配一、二、四等，與群母或云母互補。
• 來母也可以配一至四等韻母，但二等字數較少。
• 「微廢文欣元嚴凡」七個韻系只與鈍音聲母相拼。
• 董韻三等、湩韻一等及駭韻合口少字。
• 「鍾宵鹽」這幾個韻系以及藥韻，幾乎不能配莊組聲母。
• 支韻合口（舉平以賅上去入）不能配莊組聲母。
• 「之魚」韻系無唇音聲母字。
• 佳韻系齒音聲母少字。
• 祭韻 A 類開口無精組聲母字、祭韻 B 類無唇音字。
• 戈韻三等只能配牙喉音聲母；麻韻三等只能配齒音聲母。
• 曾攝合口只在牙喉音聲母有字，且字數不多。蒸拯證三個韻目完全沒有合口，但職韻目有。
• 「齊佳皆夬先陽唐庚耕清青蒸登」等韻系，開合口只在牙喉音聲母有對立。而「支脂祭泰灰咍真諄魂痕寒桓刪山仙歌戈麻」等韻系，其開合口則在舌牙齒喉音聲母都有對立。

反切

• 反切上字決定聲母，反下字決定韻母和聲調。
• 反切上下字一般都是三等或都不是三等。 如果上下字的其中一個是合口字的話，那麼這個發音一般都是合口的，但請注意唇音字的韻母不區分開合口。
• 若反切上字區分重紐，則同上字。
• 鈍音聲母反切時，上字 AB 類不互切，C 類則 ABC 三類皆可切。
• 若反切下字為 A 類、精組、章組、以母，則切音為 A 類；若反切下字為 B 類、莊組、云母，則切音為 B 類；知組及來母則沒有明顯傾向，AB 類皆可混用。

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

Introduction

Latin nouns are traditionally classified into five declensions according to their inflection patterns (adjectives generally use only the first three, while pronouns are usually irregular), as well as a few irregular nouns.

• Nominative: Indicates the subject of a sentence.
• Genitive: Indicates the processor.
• Dative: Indicates the recipient of the action, or the indirect object.
• Accusative: Indicates the object of a sentence.
• Ablative: Ablative in Latin has multiple functions:
  • Ablative of separation: 表示動作出發、離開的地方。視乎其修飾的詞彙，可能有也可能沒有介詞。抽象意義的分離不加介詞，具體意義的分離，且被修飾的詞彙不具有方位格時，添加介詞如 ab, ex, dē。
  • Ablative of place: 表示某事物所在的位置。用於條飾不具有方位格的詞彙時，添加介詞如 in, sub , prō。
  • Ablative of means/instrument: 表示行為所使用的工具。工具奪格前不加介詞。
  • Ablative of accompaniment: 介詞 cum 後加奪格表示伴隨。
• Vocative: Used to directly call on a person or thing.
• Locative: Indicates the position of a thing. Only some nouns have a locative form, other nouns express this meaning with a “preposition + ablative” structure.

General rules

• All endings of all declensions begin with a vowel, except some third declension words, and this vowel determines the word’s declension class. The differences between the various declension classes are simply the differences in these vowels (and some irregularities resulting from the evolution process). These vowels were originally part of the stem, but over a long period of evolution, they gradually merged with the ending and were subsequently reanalyzed as part of the ending.
• The nominative singular usually ends in either -s or null.
• The genitive singular and the nominative plural ends in either -ī or -s.
• The dative singular usually ends in -ī, except there a diphthong smoothing occurs in the second declension.
• Non-neuter accusative singulars end in -m, and the neuter ones are always identical to its nominative singular.
• The ablative singular ends in a vowel¹, which is a long vowel except in the third declension.
• The genitive plural ends in -um. The “long vowel + -rum” endings for the first, second and fifth declension are borrowed from pronoun paradigms.
• Non-neuter accusative plurals end in “long vowel + -s”.
• The dative/ablative plural ends in either -īs or -bus.
• The same form is used for both nominative and vocative, except in the second declension.
• The plurals of dative and ablative are identical for all nouns, with no exceptions.
• The nominative, accusative and vocative are identical for all neuter nouns, with an -a ending in their plurals, with no exceptions.

First & second declension

The stem final vowel of the first declension is a, and that of the second declension is o.

• The dative/ablative plural ending of these two declension classes were originally -ais and -ois respectively, later evolved to -īs in Classical Latin, which is a regular historical phonological process in Latin². Fīlia, fīliae and dea, deae have an irregular dative/ablative plural form ending in -ābus.
• The -us and -um endings in the second declension were -os and -om originally, the evolution to -us and -um in Classical Latin is also a regular phonological process³. The -os and -om endings are kept in monosyllabic pronouns.
• The -ō ending of the dative singular in the second declension was simplified from earlier -ōi.
• The -ī ending of the genitive singular in the second declension shuold be an irregular innovation. Before the time where -ei and -i started mixing up in Old Latin, the genitive singular was never written as -ei (even -osio was attested in Old Latin), showing that it did not come from -oi, which is different from the nominative plural.
• If a second declension noun ends in -ius or -ium in the nominative singular, its vocative singular would be -ī in Classical Latin (earlier -ie).

After the above back-reasoning process, the word forms of the two declensions will become extremely similar, differing only in the vowel at the beginning of the ending.

The second declension has a subclass, in which their nominative singular would have *-rus actually become -(e)r instead, such as ager — agrī, liber — librī, puer — puerī, vir — virī. These words have the same vocative singular and nominative singular, while the other inflections can be deduced from its genitive singular form, which is identical to all other second declension nouns.

Third declension

The third declension is a conflation of two originally distinct declension classes, one of which has a stem final vowel of i, while others have no stem final vowel. In Classical Latin, the two declension classes were almost mixed up entirely, their genitive plural remain distinct, being -um and -ium respectively.

The correspondence between the nominative singular and other word forms in the third declension has various types and is traditionally considered irregular. A decent amount of correspondence can actually be deduced from Latin’s (diachronical) sound rules, nevertheless:

Fourth & fifth declension

The stem final vowel of the fourth declension is u, and that of the fifth declension is ē.

The fourth and fifth declensions are the less used ones in Latin. Many words in these two declension classes were redistributed into the second or third declension during their development into Vulgar Latin or various Romance languages.

• The -ibus ending of the dative/ablative plural in the fourth declension came from earlier -ubus. A couple of words keep the -ubus ending, while the majority has undergone the rule of weakening word medial short vowels to i.

Irregular nouns

• Fīlia, fīliae and dea, deae have an irregular dative/ablative plural form ending in -ābus.
• The vocative singular of deus, deī is deus instead of *dee; Optional contracted forms exist for multiple hiatus-involving forms.
• Vīrus, vīrī is a second declension neuter noun with a citation form ending in -us (instead of -um). This word is an uncountable noun in Classical Latin. New Latin borrowed this word to express the meaning of “virus”, and a set of plural forms were made up by analogy to other second declension nouns.
• Domus, domūs/domī is a hybrid of second and fourth declension, with different mixing combination attested.

Adjectives

The declension of adjectives can be divided into the following categories:

• First/second declension adjectives: The masculine and neuter forms use the second declension, while the feminine forms use the first declension.
• Third declension adjectives: The nominative singular form can be further divided into three subclasses according to different genders, while other forms are similar to those of nouns in the third declension. Most (but not all) third declension adjectives have -ium in the genitive plural.
• • One termination: The nominative singular is identical in all three genders.
• • Two terminations: The nominative singular is identical in masculine and feminine, while a different form is used in neuter.
• • Three terminations: The nominative singular is different in all three genders. Usually -(e)r for masculine, -ris for feminine and -re for neuter.

Numerals

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

This page is still work in progress, stay focus for continuous upcoming updates on this page!

• 本文使用的粵語拼音系統為粵拼。
• 本文使用的閩南語拼音系統為台羅。
• 本文使用的客家語拼音系統為客語白話字。不過目前還不是很確定本站主要會使用哪種客家話拼音系統。
• 有些字在用作化學用字時的讀音，與同形古字用作古義時的讀音未必相同。
• 某部份字典查不到的讀音，是我本人的原創研究，從造字原則類推而來。

Elements

Organic Chemistry

| Symbol | English | Char | Mandarin (B) | Mandarin (P) | Cantonese | Hokkien | Hakka | Remarks | | – | – | – | – | – | – | – | – | | alkane | 烷 | ㄨㄢˊ | wán | jyun4 完, jyun2 宛 | uân | vàn, yèn | | | alkene | 烯 | ㄒㄧ | xī | hei1 稀 | hi | hî | | | alkyne | 炔 | ㄑㄩㄝ | quē | kyut3 缺 | khuat | khiet | | | alcohol | 醇 | ㄔㄨㄣˊ | chún | seon4 純 | sûn | sùn | | | aldehyde | 醛 | ㄑㄩㄢˊ | quán | cyun4 全 | tsuân | chhièn | | | ketone | 酮 | ㄊㄨㄥˊ | tóng | tung4 同 | tông | thùng | | | acyl group | 醯（台） | ㄒㄧ | xī | hei1 稀 | hi | hî | | | acyl group | 酰（港、陸） | ㄒㄧㄢ | xiān | sin1 先 | sian | siên | | | phenol | 酚 | ㄈㄣ | fēn | fan1 分 | hun | fûn | | | ether | 醚 | ㄇㄧˊ | mí | mai4 迷 | bî | mì | | | ester | 酯 | ㄓˇ | zhǐ | zi2 指、zi1 脂 | tsí | chṳ́ | | | carboxylic acid | 羧 | ㄙㄨㄛ | suō | so1 梳 | so | sô | | | acidic anhydride | 酐 | ㄍㄢ | gān | gon1 干 | kan | kôn | | | amine | 胺 | ㄢ（台）
ㄢˋ（陸） | ān（台）
àn（陸） | on1 安、on3 按 | an | ôn | | | hydrazine | 肼 | ㄐㄧㄥˇ | jǐng | zeng2 井 | tsíng | chiáng | | | nitrile | 腈 | ㄐㄧㄥ | jīng | zing1 晶、cing1 青 | tsing | chîn, chiâng | | | phosphine | 膦 | ㄌㄧㄣˊ（台）
ㄌㄧㄣˋ（陸） | lín（台）
lìn（陸） | leon4 鱗、leon6 論 | lîn, līn | lìn, lin | | | arsine | 胂 | ㄕㄣ（台）
ㄕㄣˋ（陸） | shēn（台）
shèn（陸） | san1 申、san6 腎 | sin, sīn | sṳ̂n, sṳn | | | oxime | 肟 | ㄨㄛˋ | wò | ok3 噁 | ok | ok | | | sulphone, sulfone | 碸 | ㄈㄥ | fēng | fung1 風 | hong | fûng | | | benzene | 苯 | ㄅㄣˇ | běn | bun2 本 | pún | pún | | | benzyl group | 苄 | ㄅㄧㄢˋ | biàn | bin6 卞 | piān | phien | | | naphthalene | 萘 | ㄋㄞˋ | nài | noi6 奈 | nài | noi | | | anthracene | 蒽 | ㄣ | ēn | jan1 恩 | un, in | ên | | | phenanthrene | 菲 | ㄈㄟ | fēi | fei1 非 | hui | fî, fuî | | | phenalene | 萉 | ㄈㄟˋ | fèi | fei4 肥 | huî | fì, fuì | | | chrysene | 䓛 | ㄑㄩ | qū | wat1 屈、gwat1 骨 | khut | khiut | | | pyrene | 芘 | ㄅㄧˇ | bǐ | bei2 比 | pí | pí | | | phane | 蕃 | ㄈㄢ、ㄈㄢˊ | fān, fán | faan1 番 | huan | fân | | | azulene | 薁 | ㄠˋ、ㄩˋ | ào, yù | ou3 奧、juk1 郁 | ò, hiok | au, o, yuk | | | indene | 茚 | ㄧㄣˋ | yìn | jan3 印 | ìn | yin | | | acenaphthylene | 苊 | ㄜˋ | è | aak1 厄 | ik | ak | | | fluorene | 芴 | ㄨˋ | wù | mat6 勿 | bu̍t | vu̍t | 非讀若「笏」。 | | perylene | 苝 | ㄅㄟˋ | bèi | bui6 背 | puē | poi | | | coronene | 蒄 | ㄍㄨㄢ | guān | gun1 官 | kuan | kôn | 非讀若「蔻」。 | | rubicene | 葒 | ㄏㄨㄥˊ | hóng | hung4 紅 | hông | fùng | | | cyclopentadiene | 茂 | ㄇㄠˋ | mào | mau6 貿 | bōo | meu | | | hydrocarbon | 烴 | ㄊㄧㄥ | tīng | ting1 聽 | thing | thîn, thiâng | 碳氫切 | | carbonyl group | 羰 | ㄊㄤ | tāng | tong1 湯 | thong | thông | 碳氧切 | | hydroxy group | 羥 | ㄑㄧㄤˇ | qiǎng | koeng5 鏹 | khióng | khiông | 氫氧切 | | thiol | 巰 | ㄑㄧㄡˊ | qiú | kau4 求 | k(h)iû | khiù | 氫硫切 | | | | | | | | | |

Miscellaneous

| Symbol | English | Char | Mandarin (B) | Mandarin (P) | Cantonese | Hokkien | Hakka | Remarks | | – | – | – | – | – | – | – | – | | ammonium | 銨 | ㄢ（台）
ㄢˇ（陸） | ān（台）
ǎn（陸） | on1 安 | an | ôn | | | phosphonium | 鏻 | ㄌㄧㄣˊ | lín | leon4 鱗 | lîn | lìn | | | arsonium | 鉮 | ㄕㄣˊ | shén | | sin | sṳ̂n | | | ammonia | 氨 | ㄢ | ān | on1 安 | an | ôn | | | cyanide | 氰 | ㄑㄧㄥ（台）
ㄑㄧㄥˊ（陸） | qīng（台）
qíng（陸） | cing1 青、cing4 晴 | tshing, tsîng | chhîn, chhiâng, chhìn, chhiàng | | | | | | | | | | |

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

Finally get a self-hosted blog (somewhat) set up online, but still struggling to get everything working.

Since I’m not quite familiar with the framework, there may be a lot of bugs around the site. So feel free for providing technical supports by contacting me or issue PRs directly on the GitHub repo! Thank you for the support from all the readers!

📌 Update on 2026/03/14

After playing around with various settings, most of the visible UI issues have finally been resolved! Though taking longer than I initially expected. However, I still don’t figure out why the “reading time” isn’t displaying at all.

📌 Update on 2026/04/13

Major UI revamp. The not-displaying “reading time” issue has been resolved.

📌 Update on 2026/04/23

Change the colour scheme on the entire site.

📌 Update on 2026/05/29

Updated the entire website’s attitude towards CCP’s propaganda channel Xu Xiaoxiang, and forcefully changed all titles and content quoted from him into simplified Chinese.

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

Alex Hsu recently published two posts in a row discussing the topic of regional variations in Chinese.

• 《The tale of two translation houses》
• 《How Taiwan and China rename English movies (and why it’s a mess)》

Admittedly, from the perspective of “movie title translations,” the discrepancies between translations by distributors in different regions are particularly severe (especially some poorly translated ones, Taiwan seems to favor using “formulas” or even making dirty jokes). Alex has suggested that “film studios should unify their translations”. However, direct translation isn’t always the best approach, since different languages ​​(and even different regions speaking the same language) can have different cultures, and direct translation may not always allow audiences from certain cultures to grasp the film’s theme or the artistic conception of the original language. (Some countries also have a tendency to include spoilers when translating foreign film titles) However, blindly forcing uniformity could also backfire.

Perhaps we should leave the previously established (or “accustomed to mistakes”) names for shit mountain untouched. It would be better to focus on choosing good names for the newly emerging ones.

Annihilating local languages?

I really admire what The Pokemon Company did. In 2016, they unified twenty years of divergent names. Taiwan had been calling it 《神奇寶貝》(“Magical Treasures”), Hong Kong used 《寵物小精靈》(“Pet Elves”), and China called it 《口袋妖怪》(“Pocket Monsters”). They scrapped all three and unified under one name: 《寶可夢》(“Pokémon,” phonetically). I remember thinking the new name sounded terrible. But I agreed with the decision. And time proved them right.

——Alex Hsu《How Taiwan and China rename English movies (and why it’s a mess)》

Pokémon is an example of an attempt to clean up the shit mountains. Alex’s remarks actually overlooked a very important aspect. When Pokémon (and other Nintendo products) were unifying their transliterations, they often tilted at Mandarin pronunciations, leading to accusations of “belittling Cantonese and substituting Mandarin for Cantonese”, which caused a stir in Hong Kong at the time. If you were to say something like that in a Hong Kong community, you’d very likely get a lot of flak. If I had a less favourable impression of Alex, I could immediately launch into a tirade, accusing him of being a Mandarin chauvinist (I really could do that), willing to sacrifice the rationality of local dialect¹ pronunciation and completely deny the value of local dialects in order to achieve his grand unification ideal.

寶可夢就是一個嘗試清理屎山的例子。Alex 的這番言論其實故意忽略了一個非常重要的面向。 寶可夢（以及其他任天堂旗下產品）在制定統一譯名的音譯時往往以官話發音為主，導致其被指責「矮化粵語、以普代粵」，當時在香港炎上過一陣子。你這番言論要是挪到香港人的圈子裏頭講，有很高機率會被罵死。因此我立馬就可以開罵攻擊 Alex，罵牠是支持官話霸權、罵牠是官話沙文主義者、罵牠是支共喉舌，會為了完成牠的大一統理念，而不惜犧牲地方語言¹發音的合理性，並全盤否定地方語言的存在價值，與支共實乃同一陣線。誰採信牠的言論，誰就是匪諜，應該徹底剷除。

• 妈的，你是不是中国人，是中国人就该讲普通话，别唧唧歪歪讲什么鸟语！
• 不说普通话的是不是辱华废青啊？
• 讲什么狗屁鸟语，等着被统一吧！
• 咱们有推出官方汉化版本，你们就得感恩了！
• XX 人全都是二鬼子，一句普通话不说

——Design dialogue

However, on the other hand, some translations of names made in Hong Kong are also poorly done. Many translations in the sports world are based on the Cantonese pronunciation, but they often misjudge the original pronunciation, and the resulting translations are almost like random translations.

The Romanization school of thought

You say you’re not Chinese? Then don’t use Chinese characters!

——Romanized Taiwanese chauvinists

Tan Kian-Tiong mentioned in his article “葬送的芙莉蓮 EP1-2 語句選”, that he persuaded the team at PTS Taigi Channel to write transliterated foreign words in Latin letters (instead of directly borrowing characters from Mandarin Chinese), arguing that the Ministry of Education’s Dictionary of Frequently-Used Taiwanese Taigi also uses this format.

Such a mix of Latin characters is not a convention in written Chinese in Hong Kong, except in certain compromises (such as the case of “Pokémon” mentioned above).

The impossible triangle of faithfulness, expressiveness, and elegance

One compromise solution that takes all the above arguments into account, is to try to find characters that sound similar in various Chinese dialects¹ when creating new transliterations (unless they deviate too much from established conventions or “accustomed to mistakes”). However, this adds many more issues for translators to consider:

• Faithfulness: Does it accurately convey the pronunciation of the original text?
• Expressiveness: Does it accurately convey the meaning and artistic conception of the original text?
• Elegance: Will the combined translation sound unrefined (in some dialects)?

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

Recently, several articles discussing vocational high schools (高職) in Taiwan have appeared on “Wiwi Blog Universe”:

• Eo《後悔讀高職》
• JN《讀高職是我一點都不後悔的事情》
• ikuka《讀高職不代表我成績差》
• 黃碩人《高職 FAQ》

Let me also tell you about Hong Kong’s education system.

In Hong Kong, most mainstream secondary schools allow students to choose their elective subjects for their senior secondary education. After choosing your subjects, you don’t need to ever touch the subjects you didn’t choose again, besides Chinese, English, Mathematics, and Liberal Studies (or later “Citizenship and Social Development”) and two or three elective subjects. So I really didn’t have to worry about what to do with the humanities subjects that I wasn’t good at in my junior secondary.

… Unless the gap in your grades lies in your core subjects. My own situation is that I’m quite good at maths, okay at English, and Chinese and Liberal Studies are the worst, and my grades in maths and various science electives aren’t particularly outstanding either (didn’t get those “star grades” in DSE), which is quite a rare type, and can’t find a place within the Hong Kong education system¹. So after failing Chinese in DSE, I was forced to take English continuing education courses and higher diploma programmes with countless others who failed English instead, even though my original English proficiency wasn’t too bad², which is quite absurd.

⚠️ Notice

This post is a draft translation from the Chinese version which have not yet been thoroughly proofread.

My blog currently has three languages: written Chinese, Cantonese, and English. The different language versions are currently distinguished by different paths.

Since I rarely promote this website in circles where people don’t understand Chinese, most of the English content on this site has been roughly translated using machine translation tools, and I’ll proofread it more carefully later when I have time.

JN recently created a tool called IndiePing, which can be used to search for posts in the “BlogBlog Club Universe” that link to your blog. The tool has a minor bug: if your blog is not under the root directory of a subdomain, IndiePing might retrieve the wrong blog name. JN replied that this bug needs to be manually fixed case by case. Currently, this blog doesn’t have its own domain, but perhaps one day I’ll move the entire blog to my own domain. I wonder if, when moving the domain in the future, the three different language versions should maintain directory distinctions, or if, for compatibility reasons with some blog linking tools, I’ll have to assign different hostnames to each language version.

Additionally, noeFly is also developing a blog search engine, but it’s unclear how well it supports non-root directory + multilingual blogs like mine.

Eddie Lv somehow decided to include a link to my English website for some reason on his “Mentions & Discussion” page, even though Eddie’s website is written in Chinese. Perhaps he doesn’t notice that my website actually has a Chinese version.