search

jdlorimer / chinese-support-redux

Anki add-on providing support for Chinese study
https://ankiweb.net/shared/info/1128979221
GNU General Public License v3.0
101 stars 50 forks source link

Some characters get assigned incorrect readings randomly #173

Open soliviantar opened 3 years ago

soliviantar commented 3 years ago

First of all, thanks for your great work on thie addon! It is a life saver.

Describe the bug For some reason, some characters like 说 seem to get assigned a reading... randomly. Sometimes it gets a shuō reading and sometimes the shuì reading. I can't find a way to control which one it'll get. I suppose 说 should have the default set to shuō in sentences like 你跟他说, 什么说呀, 把话说清楚, 说一下你的地址 or 说吧. But I am getting shuì for these. I can run a search and replace for the shui's and change it for shuo's and running bulk-fill transcriptions again will change the color of the pinyin, but not of the Color field. I have to manually go to the Reading field and tab away from it to make it update the Color field. This happened in about 170 cards out of 274 having 说. The dū reading also appears for 都 in all cases, I think. But that is easier to fix since it doesn't affect tone.

To Reproduce Steps to reproduce the behavior: Add sentence with character 说 Bulkfill transcriptions Some 说 will get shui reading and be colored as 4th tone Changing the tone in the Reading field changes the reading color, but not the hanzi color

Expected behavior I think the shuō reading should be the default for most cases.

Specs (please complete the following information):

Something similar was reported here: https://github.com/luoliyan/chinese-support-redux/issues/144

jdlorimer commented 3 years ago

It's not random, exactly. I'm pretty sure I know why it's happening - it's pulling all the dictionary entries into an array and then selecting the first reading if there's more than one.

It shouldn't happen, but I've never settled on a strategy for dealing with ambiguous entries. The solution will probably involve using some frequency data to weight the choices.

I'll take another look at this soon.

noinkling commented 3 years ago

I think you need to remove/exclude/bypass single-character words from the word dictionary. That way it can fall back to the individual hanzi data (which provides the most common reading).

noinkling commented 3 years ago

There are also some multi-character words with multiple readings, so if you wanted to be thorough you could try and get some reading frequency data for those and/or use some kind of heuristic to pick the best one.

Here's a query that can help examine:

SELECT * FROM (
    SELECT *, count(*) OVER (PARTITION BY simplified) AS count
    FROM cidian c
    WHERE length(simplified) > 1
)
WHERE count > 2
ORDER BY simplified, traditional, pinyin;

Note that a lot of those rows have the same pinyin reading, just minor differences in notation format, so you/we might want to improve on the normalization in the update script so they're considered equal.

joeminicucci commented 3 years ago

@noinkling

There are also some multi-character words with multiple readings, so if you wanted to be thorough you could try and get some reading frequency data for those and/or use some kind of heuristic to pick the best one.

Here's a query that can help examine:

SELECT * FROM (
  SELECT *, count(*) OVER (PARTITION BY simplified) AS count
  FROM cidian c
  WHERE length(simplified) > 1
)
WHERE count > 2
ORDER BY simplified, traditional, pinyin;

Note that a lot of those rows have the same pinyin reading, just minor differences in notation format, so you/we might want to improve on the normalization in the update script so they're considered equal.

Based on the data available in ccedict, I do not believe there is a current heuristic available with what we currently have. Is there another data source we can pull to check lexical usage frequency?

chambm commented 2 years ago

I ran into this problem as well. Finding comprehensive frequency data based on pronunciation may be very difficult. Instead, or in the meantime, we could add a manually curated file that sets overrides for common polyphonic words and characters. Shou1, dou1, he2, etc. 啊 has literally every tone in chinese.db (a, a1, a2, a3, a4, a5)! The manually curated file would pick a canonical pinyin for each character/word and we could also allow users to override that.

chambm commented 2 years ago

This is a hard problem. There are quite a few common polyphones that are commonly used with multiple pronunciations. And even Google hasn't really got it figured out. On Google Translate, 行行 gets pronounced as "xing2xing2" but the pinyin transcription is "hang2hang2"! But apparently it should be hang4hang4 according to https://zd.hwxnet.com/search.do?keyword=%E8%A1%8C&x=0&y=0 .

My wife is Chinese so I tasked her with picking the most common pronunciation for a bunch of single character polyphones. It was rather amusing to watch her balk at the task when I gave her the list without accounting for frequency. In English we have tons of rarely used words, but they are usually long and unambiguous (but often difficult) to pronounce. Not so in Chinese, apparently. After I sorted the list by frequency she got a lot of them done and just left a few that she insists are pretty even:

hanzi most common alt1 alt2 alt3 alt4 alt5
处   chu3 chu4      
行   hang2 xing2      
数   shu3 shu4 shuo4    
干   gan1 gan4      
量   liang2 liang4      

There are more she left blank but I'm not sure if she just hasn't got to them yet (they're all over rank 2000 in frequency).

Here are her choices for the others: hanzi most common alt1 alt2 alt3 alt4 alt5
的 de5 de5 di1 di2 di4  
和 he2 he2 he4 hu2 huo2 huo4
了 le5 le5 liao3 liao4    
为 wei4 wei2 wei4      
将 jiang1 jiang1 jiang4 qiang1    
说 shuo1 shui4 shuo1      
与 yu3 yu2 yu3 yu4    
上 shang4 shang3 shang4      
大 da4 da4 dai4      
要 yao4 yao1 yao4      
地 de5 de5 di4      
据 ju4 ju4 ju1      
着 zhe5 zhao1 zhao2 zhe5 zhuo2  
种 zhong3 zhong3 zhong4      
把 ba4 ba3 ba4      
比 bi3 bi3 bi1      
好 hao3 hao3 hao4      
同 tong2 tong2 tong4      
分 fen1 fen1 fen4      
更 geng4 geng1 geng4      
会 hui4 hui4 kuai4      
可 ke3 ke3 ke4      
几 ji3 ji1 ji3      
给 gei3 gei3 ji3      
场 chang3 chang2 chang3      
占 zhan4 zhan4 zhan1      
得 de2 de2 de5 dei3    
长 chang2 chang2 zhang3      
作 zuo4 zuo1 zuo4      
阿 a1 a1 e1      
号 hao4 hao2 hao4      
只 zhi3 zhi3 zhi1      
看 kan4 kan1 kan4      
正 zheng4 zheng1 zheng4      
强 qiang2 jiang4 qiang3 qiang2    
间 jian1 jian1 jian4      
卡 ka3 ka3 qia3      
当 dang1 dang1 dang4      
打 da3 da2 da3      
底 di3 de5 di3      
夫 fu1 fu1 fu2      
称 cheng1 chen4 cheng1 cheng4    
便 bian4 bian4 pian2      
头 tou2 tou2 tou5      
那 na4 na3 na4      
少 shao3 shao3 shao4      
难 nan4 nan2 nan4      
发 fa1 fa1 fa4      
令 ling4 ling2 ling3 ling4    
重 zhong4 chong2 zhong4      
创 chuang4 chuang4 chuang1      
没 mei2 mei2 mo4      
率 shuai4 lü4 shuai4      
奇 qi2 ji1 qi2      
塞 sai1 sai1 sai4 se4    
华 hua2 hua2 hua1      
勒 le4 le4 lei1      
度 du4 du4 duo2      
差 cha1 cha1 cha4 chai1    
边 bian1 bian1 bian5      
化 hua4 hua1 hua4      
通 tong1 tong1 tong4      
合 he2 ge3 he2      
跑 pao3 pao2 pao3      
菲 fei1 fei1 fei3      
见 jian4 jian4 xian4      
降 jiang4 jiang4 xiang2      
子 zi3 zi3 zi5      
扎 zha1 zha1 zha2 za1    
草 cao3 cao3 cao4      
远 yuan3 yuan3 yuan4      
供 gong1 gong1 gong4      
耶 ye1 ye1 ye2 ye5    
什 shen2 shen2 shi2      
朝 chao2 chao2 zhao1      
假 jia3 gei1 jia3 jia4    
听 ting1 ting1 ting4      
尽 jin4 jin3 jin4      
呢 ne5 ne5 ni2      
藏 cang2 zang4 cang2      
转 zhuan3 zhuai3 zhuan3 zhuan4    
蒙 meng2 meng2 meng1 meng3    
价 jia4 jia4 jie5      
板 ban3 ban3 pan4      
调 diao4 diao4 tiao2      
压 ya1 ya1 ya4      
宁 ning2 ning4 ning2      
句 ju4 gou1 ju4      
抢 qiang3 qiang1 qiang3      
足 zu2 ju4 zu2      
空 kong1 kong1 kong4      
论 lun4 lun2 lun4      
卷 juan3 juan3 juan4      
吗 ma5 ma3 ma5      
传 chuan2 chuan2 zhuan4      
般 ban1 ban1 pan2      
雨 yu3 yu3 yu4      
追 zhui1 dui1 zhui1      
脚 jiao3 jiao3 jue2      
校 xiao4 jiao4 xiao4      
圈 quan1 juan1 juan4 quan1    
娜 na4 na4 nuo2      
落 luo4 la4 lao4 luo4    
倒 dao3 dao3 dao4      
伯 bo2 ba4 bai3 bo2    
结 jie2 jie1 jie2      
冲 chong1 chong1 chong4      
待 dai4 dai1 dai4      
鸟 niao3 diao3 niao3      
采 cai3 cai3 cai4      
读 du2 dou4 du2      
哪 na3 na3 na5 nei3    
杆 gan1 gan1 gan3      
丽 li2 li2 li4      
氏 shi4 shi4 zhi1      
堡 bao3 bao3 pu4      
色 se4 se4 shai3      
畜 chu4 chu4 xu4      
铺 pu4 pu4 pu1      
佛 fo2 fo2 fu2      
炸 zha4 zha2 zha4      
免 mian3 mian3 wen4      
答 da2 da1 da2      
骑 qi2 ji4 qi2      
载 zai3 zai3 zai4      
喝 he1 he1 he4      
幢 zhuang4 chuang2 zhuang4      
背 bei1 bei1 bei4      
页 ye4 xie2 ye4      
奔 ben1 ben1 ben4      
症 zheng4 zheng4 zheng1      
横 heng2 heng2 heng4      
食 shi2 shi2 si4      
尾 wei3 wei3 yi3      
稍 shao1 shao1 shao4      
术 shu4 shu4 zhu2      
语 yu3 yu3 yu4      
累 lei4 lei3 lei4      
趟 tang4 tang1 tang4      
亲 qin1 qin1 qing4      
划 hua2 hua2 hua4      
涌 yong3 chong1 yong3      
散 san4 san4 san3      
担 dan4 dan1 dan4      
赚 zhuan4 zhuan4 zuan4      
服 fu2 fu2 fu4      
操 cao1 cao1 cao4      
解 jie3 jie3 jie4      
咱 za2 zan2 za2      
岭 ling3 ling2 ling3      
鲜 xian1 xian3 xian1      
弹 tan2 dan4 tan2      
涨 zhang3 zhang3 zhang4      
甚 shen4 shen2 shen4      
薄 bao2 bao2 bo4      
撒 sa3 sa1 sa3      
斗 dou3 dou3 dou4      
禁 jin4 jin1 jin4      
泥 ni2 ni2 ni4      
挑 tiao1 tiao1 tiao3      
埋 mai2 mai2 man2      
钻 zuan4 zuan4 zuan1      
折 zhe2 she2 zhe1 zhe2    
监 jian1 jian1 jian4      
挣 zheng4 zheng1 zheng4      
挡 dang3 dang3 dang4      
摸 mo1 mo1 mo2      
炮 pao4 bao1 pao2 pao4    
参 can1 can1 shen1      
辟 pi4 bi4 pi4      
似 si4 shi4 si4      
屯 tun2 tun2 zhun1      
虾 xia1 ha2 xia1      
弄 nong4 long4 nong4      
秘 mi4 mi4 bi4      
喷 pen1 pen1 pen4      
荷 he2 he2 he4      
泡 pao4 pao1 pao4      
浅 qian3 jian1 qian3      
否 fou3 fou3 pi3      
混 hun4 hun2 hun4      
匹 pi3 pi1 pi3      
磨 mo2 mo2 mo4      
沈 shen3 chen2 shen3      
模 mo2 mo2 mu2      
夹 jia2 jia1 jia2 jia4    
踏 ta4 ta1 ta4      
渐 jian4 jian1 jian4      
尺 chi3 che3 chi3      
扛 kang2 gang1 kang2      
喂 wei4 wei2 wei4      
浆 jiang1 jiang1 jiang4      
轴 zhou2 zhou2 zhou4      
谜 mi2 mei4 mi2      
档 dang3 dang3 dang4      
拚 pin1 pan4 pin1      
脏 zang1 zang4 zang1      
歪 wai1 wai1 wai3      
莎 sha1 sha1 suo1      
扫 sao3 sao3 sao4      
沉 chen2 chen1 chen2      
识 shi2 shi2 zhi4      
侧 ce4 ce4 zhai1      
仔 zai3 zai3 zi1 zi3    
恶 e4 e3 e4 wu4    
晃 huang4 huang3 huang4      
尿 niao4 niao4 sui1      
臭 chou4 chou4 xiu4      
饮 yin3 yin3 yin4      
柜 gui4 gui4 ju3      
脉 mai4 mai4 mo4      
芯 xin1 xin1 xin4      
凉 liang2 liang2 liang4      
妻 qi1 qi1 qi4      
挨 ai1 ai1 ai2      
缝 feng2 feng2 feng4      
腊 la4 xi1 la4      
刹 sha1 cha4 sha1      
罢 ba4 ba4 ba5      
宿 su4 su4 xiu3 xiu4    
刷 shua1 shua1 shua4      
俊 jun4 jun4 zun4      
膏 gao1 gao1 gao4      
抹 mo3 ma1 mo3 mo4    
削 xiao1 xiao1 xue1      
唯 wei2 wei2 wei3      
咋 za3 za3 ze2 zha4    
旋 xuan2 xuan2 xuan4      
钉 ding1 ding1 ding4      
拾 shi2 she4 shi2      
澄 cheng2 cheng2 deng4      
掺 chan1 chan1 shan3      
桔 ju2 jie2 ju2      
杠 gang4 gang1 gang4      
锯 ju4 ju1 ju4      
厕 ce4 ce4 si4      
匾 bian3 bian3 pian2      
揣 chuai3 chuai1 chuai3      
弟 di4 di4 ti4      
笼 long2 long2 long3      
衰 shuai1 cui1 shuai1      
胖 pang4 pan2 pang4      
汞 gong3 gong3 hong4      
漂 piao1 piao1 piao3 piao4    
殷 yin3 yan1 yin3      
劈 pi1 pi1 pi3      
驮 tuo2 tuo2 duo4      
刨 pao2 bao4 pao2      
粥 zhou1 yu4 zhou1      
朴 pu3 po4 pu3      
估 gu1 gu1 gu4      
熬 ao2 ao1 ao2      
隐 yin3 yin3 yin4      
卒 zu2 zu2 cu4      
么 me5 ma2 ma5 me5    
叉 cha1 cha1 cha2 cha3    
舌 she2 ji1 she2      
攒 zan3 cuan2 zan3      
楞 leng4 leng2 leng4      
雀 que4 qiao1 que4      
荡 dang4 dang4 tang4      
淋 lin2 lin2 lin4      
缉 ji1 ji1 qi1      
凹 ao1 ao1 wa1      

So does it seem reasonable to put these in a table in the addin for it to use to look in first for transcription before falling back on the first one in the chinese.db? I can also do two character polyphones.