mirror of
https://github.com/GrammaticalFramework/gf-core.git
synced 2026-04-13 23:09:31 -06:00
test results and documentation for Chinese
This commit is contained in:
aarne
@@ -1,322 +1,29 @@
|
||||
Chinese resource grammar experiment
|
||||
Chinese resource grammar
|
||||
|
||||
(c) Aarne Ranta 2012
|
||||
(c) Aarne Ranta, Zhuo Lin Qiqige, Qiao Haiyan 2012
|
||||
|
||||
Idea: bootstrap a complete resource grammar by
|
||||
- cloning files from another language (Thai)
|
||||
- extracting lexicon from available sources (Swadesh, HSK list, Google translate)
|
||||
- fixing the errors in syntax and lexicon
|
||||
- testing the grammar in applications
|
||||
|
||||
With next to now knowledge of Chinese, but access to web sources, a grammar book
|
||||
19/10/2012
|
||||
|
||||
Claudia Ross and Jing-heng Sheng Ma,
|
||||
Modern Mandarin Chinese Grammar. A Practical Guide,
|
||||
Routledge,
|
||||
London and New York,
|
||||
2006.
|
||||
|
||||
and an extended mini resource by Jolene Zhuo Lin qiqige
|
||||
and comments on that from Inari Listenmaa.
|
||||
This is a compilable, almost complete, but not yet perfect implementation of the GF resource grammar for Chinese.
|
||||
|
||||
The question is how good it will be before the visit to Shanghai in two weeks,
|
||||
and how much work will be needed to fix everything. But so far I feel like the
|
||||
guy in Searle's "Chinese Room", http://en.wikipedia.org/wiki/Chinese_room
|
||||
The Synopsis test set has the following statistics:
|
||||
|
||||
GOOD 295 68%
|
||||
SOSO 66 15%
|
||||
BAD 76 17%
|
||||
|
||||
5/10/2012
|
||||
More details:
|
||||
|
||||
Clone chinese/*Chi.gf from thai/*Tha.gf by
|
||||
|
||||
runghc lib/src/Clone.hs Tha Chi
|
||||
|
||||
This compiles directly, omitting Lexicon and Structural but producing some Thai words, visible with
|
||||
|
||||
pg -words
|
||||
|
||||
Then port parts of examples/extmini/*Cmn.gf appending them to Lexicon, Structural, Paradigms, Res.
|
||||
Tweak until everything compiles. For simplicity, retain lincat's of Thai.
|
||||
|
||||
Then identify Thai words with 'pg -words', and locate them with 'ma'. Replace them with constants
|
||||
in ResChi, initialized with defaultStr = "". 17 such constants are needed. StringsChi can be eliminated.
|
||||
It was a transient part of Tha anyway.
|
||||
|
||||
Copy examples/numerals/chinese.gf to NumeralChi.gf.
|
||||
|
||||
We now have 47 words (Chinese characters),
|
||||
|
||||
Lang> pg -words
|
||||
0 1 2 3 4 5 6 7 8 9 万 个 他 他们 仟 伍 你 你们 佰 叁 壹 壹万 壹拾
|
||||
壹拾壹 大 女人 她 好奇 小 我 我们 房子 拾 拾万 拾壹 拾壹万 捌 柒 树 棵
|
||||
每 玖 男人 睡 知道 约翰 绿 肆 贰 走 这 那 间 陆 零 非常
|
||||
|
||||
and thousands of linearizable Cl. With Thai word order so far. And all these blank (defaultStr)
|
||||
function words.
|
||||
|
||||
男人 绿 大 非常 每 男人 绿 走 大
|
||||
every green man very bigly is being walking bigly
|
||||
|
||||
约翰 睡
|
||||
John sleeps
|
||||
|
||||
约翰 睡
|
||||
John is sleeping
|
||||
|
||||
他 知道 约翰 走
|
||||
he knows that John would have walked
|
||||
|
||||
每 走
|
||||
everyone walks
|
||||
|
||||
你们 知道 走
|
||||
you know that one will walk
|
||||
|
||||
房子 我们 绿 男人
|
||||
greenest house we is being a man
|
||||
|
||||
这 走 小
|
||||
this is being walking smally
|
||||
|
||||
Ca. 2 hours of work has gone to all this.
|
||||
|
||||
Next step: Chinese Swadesh list, http://en.wiktionary.org/wiki/Appendix:Mandarin_Swadesh_list
|
||||
|
||||
-- № English word POS Pinyin IPA notes Traditional Chinese Simplified Chinese
|
||||
|
||||
> let analyse line = case words line of n:eng:ws | all Data.Char.isDigit n -> eng ++ " " ++ last ws ; _ -> ""
|
||||
> readFile "swadesh.txt" >>= mapM_ (putStrLn . analyse) . lines
|
||||
|
||||
After most of this is done, we have 218 words:
|
||||
|
||||
一些 万 丈夫 个 云 什么 什么时候 他 他们 仟 伍 你 你们 佰 光滑 冰 冷 刺 割 劈开
|
||||
动物 厚 叁 叶 吃 名 吐 听 吵架 吹 呕 呼吸 和 咬 哪里 唱 喝 嗅 嘴 因为 圆 在...里
|
||||
地球 坏 坐 壹 壹万 壹拾 壹拾壹 多 夜晚 大 天 太阳 头 头发 女人 她 好 好奇 如何 如果
|
||||
妻子 孩子 宽 对 小 少 尖 尘土 尾 山 干 年 心脏 怕 想 我 我们 房子 手 打 打猎 扔 拉
|
||||
拾 拾万 拾壹 拾壹万 指甲 挖 挤 捌 推 揉 握 搔 擦 数 新 星 月亮 杀 来 柒 树 树枝 树皮
|
||||
根 森林 棵 死 每 水 水果 沙 河 洗 活 流 浮 海 温暖 游泳 湖 湿 满 火 灰 烂 烟 爱 牙齿
|
||||
狗 玖 玩 男人 白 白天 皮 盐 直 看 眼睛 睡 知道 短 石 种子 窄 站 笑 红 约翰 绑 结冰 绳
|
||||
绿 缝 羽毛 翼 老 耳朵 肆 肉 肚子 肝 肠子 背 胸 脂肪 脏 脖子 脚 腿 膝 膨胀 舌 花 草 落下
|
||||
薄 虫 虱 蛇 蛋 血 角 谁 贰 走 路 躺 转 近 这 这里 那 那里 重 钝 长 间 陆 雨 雪 零 雾
|
||||
非常 风 飞 骨 鱼 鸟 黄 黑 鼻
|
||||
|
||||
We have more than a half left:
|
||||
|
||||
Lang> pg -missing -lang=Chi | ? wc -w
|
||||
329 _tmpi
|
||||
|
||||
But we get more interesting sentences:
|
||||
|
||||
Lang> gr -number=8 PredVP ? ? | l
|
||||
白天 满 冷 多 个 好奇 你们 谁
|
||||
many cold fuller days wonder who you weren't
|
||||
|
||||
胸 黄 红 一些 胸 数
|
||||
some red breast yellowly is counted
|
||||
|
||||
头 握 多 头 看
|
||||
many heads to be held see themselves
|
||||
|
||||
丈夫 那里 对 他 擦
|
||||
his husbands there correctly are wiping them
|
||||
|
||||
约翰 躺
|
||||
John lies
|
||||
|
||||
地球 这里 非常 刺 知道 我们 站
|
||||
earth very here stabbed knows that we wouldn't have stood
|
||||
|
||||
多 那里 那里 想 在...里 她
|
||||
many there there are being thinking in her
|
||||
|
||||
尘土 每 个 那里 握
|
||||
every dust there holds itself
|
||||
|
||||
Ca. 30 minutes for this phase, 2h30 total.
|
||||
|
||||
Now let's take the missing words and try to look them up the in HSK1 word list.
|
||||
Now we have 218 left, less than 200 really missing. But potentially some junk.
|
||||
|
||||
Some more words found in Google translate. Guessing mkVA and other complex subcats.
|
||||
Now we have 126 left, less than 100 really missing. pg -words shows 404 Chinese words:
|
||||
|
||||
…之间 一 一些 万 丈夫 上边 下 不 丢 个 中 为了 为什么 书 买 云 人 什么 什么时候
|
||||
今天 从 从前 他 他们 仟 以后 件 伍 会 但是 你 你们 佰 做 光滑 公寓 写 冰 冰箱 冷
|
||||
刺 前边 割 劈开 动物 医生 卖 厚 原则上 去 叁 发现 变 只 台 右 叶 号 吃 名 吐 吗
|
||||
听 吵架 吸 吹 呕 呼吸 和 和平 咬 哪里 唱 商店 啤酒 喜欢 喝 嗅 嘴 回答 因为 国 国王
|
||||
圆 在 在...里 地板 地毯 地球 坏 坐 块 塑料制成 壹 壹万 壹拾 壹拾壹 多 夜晚 大 大学
|
||||
天 天花板上 太 太阳 头 头发 女人 女王 奶 奶酪 她 好 好奇 如何 如果 妹妹 妻子 姑娘
|
||||
婴儿 存在 学 学校 学生 孩子 它 宗教 宽 寄 对 小 少 尖 尘土 尾 屋顶 山 工业 工厂
|
||||
左 已经 市 希望 帽子 干 干净 年 年轻 开 弟弟 往 心脏 必须 忘 怕 想 愿意 懂 我
|
||||
我们 或者 战 房子 所 手 手套 打 打猎 扔 报 拉 拾 拾万 拾壹 拾壹万 指甲 挖 挤
|
||||
捌 推 揉 握 搔 摄像头 摆 擦 收音机 敌人 教 教堂 数 新 旅行 时间 星 星球 是 是否
|
||||
最 月亮 有 朋友 木 机 杀 村庄 来 柒 树 树枝 树皮 根 桌子 森林 棵 椅子 次 歌 死
|
||||
母亲 每 比 水 水果 汽车 沙 河 油 洗 活 流 浮 海 海港 温暖 游泳 湖 湿 满 漂亮 火
|
||||
火车 灯 灰 炉子 烂 烟 烧 热 爱 父亲 牙齿 牛 牧师 状物 狗 猫 玖 玩 玩儿 现在
|
||||
电视 男人 男孩 画 白 白天 的 的橡胶 的距离 皮 皮革 盐 直 看 眼睛 睡 知道 短
|
||||
石 石头 离 种 种子 科学 窄 窗 站 笑 笔 等 红 约翰 纸 绑 结冰 给 绳 绿 缝 羊
|
||||
羽毛 翼 老 老师 老板 耳朵 肆 肉 肚子 肝 肠子 背 胸 脂肪 脏 脖子 脚 腿 膝 膨胀
|
||||
自己 自行车 舌 船 艺术 花 花园 苹果 草 落下 蓝 薄 虫 虱 虽然 蛇 蛋 血 衣服 表弟
|
||||
衬衫 袜子 被 角 警察 计算机 语法 语言 说 读 谁 贰 赢 走 跑 路 跳 躺 转 近 还是
|
||||
这 这里 远 道理 那 那里 酒 重 重要 金 钝 钢 铁 银 银行 长 门 问 问题 间 陆 除了…
|
||||
以外 雨 雪 零 雾 非常 面包 鞋 音乐 风 飞 飞机 饭店 马 骨 鱼 鸟 黄 黄油 黑 鼻
|
||||
|
||||
|
||||
4h work for a compiling grammar with a decently sized lexicon. The next thing
|
||||
is to fix the worst bugs, in particular in word order.
|
||||
|
||||
|
||||
6/10
|
||||
|
||||
Started a test set for syntax, with 54 sentences testing predication and noun
|
||||
phrase formation. The first run, baseline.txt, uses the Thai-based syntax.
|
||||
|
||||
As the first thing, revised the order of determiners in an NP, and the place of
|
||||
adverbs in VP. Det lincats seem to be too rich, whereas Cl may need to be made
|
||||
discontinuous, to enable the proper place of IAdv. This is correct in Jolene's
|
||||
code.
|
||||
|
||||
3h of work today, total 7h.
|
||||
|
||||
|
||||
7/10
|
||||
|
||||
Ported Jonele's lincat's everywhere, eliminated Thai-style identifiers. Difficult
|
||||
to choose default tenses. But the code is nice to work with.
|
||||
|
||||
4h today, total 11h.
|
||||
|
||||
|
||||
8/10
|
||||
|
||||
Refactored clause building with an overloaded ResChi.mkClause. Added existentials with the verb
|
||||
you_s. Very uncertain on many things, time to call an expert.
|
||||
|
||||
1h today, total 12h.
|
||||
|
||||
Following a suggestion by Thomas Hallgren, divided multicharacter words to as many tokens as
|
||||
there are characters. This means all division into tokens is performed by the parser, which in a
|
||||
sense is optimal. To "remove" spaces in linearization, simply use l -unchars; to "insert" spaces
|
||||
in parsing, use pt -chars | p. All this is done by the oper ResChi.word, which can be changed
|
||||
to change this behaviour.
|
||||
|
||||
|
||||
12/10
|
||||
|
||||
Went through some open questions with Jolene. Then fixed some parameters for Det and Adv and prepared
|
||||
Lexicon and Structural for her inspection and completions.
|
||||
|
||||
5h my time today, 3h Jolene's, total 20h.
|
||||
|
||||
|
||||
14/10
|
||||
|
||||
Lexicon and Structural checked and completed by Jolene. Some open issues found by Jolene, hopefully to be fixed this week:
|
||||
- generics for weather -> Extra
|
||||
- "something" -> Extra (stg for event vs. physical object)
|
||||
- relative clause with complex RP -> postpone; discontinuous?
|
||||
- adjectives for animates -> leave to semantics (cf. Thai)
|
||||
- weather verbs rather VP's -> leave to semantic lexicon
|
||||
- markers 把(marker for direct-object), 被(marker for indirect-object) -> fix for V2A, V3 with ba
|
||||
- positions of the parts of Subj -> fix by introducing particles before or after the subject of the main clause
|
||||
- the three de particles: 的，地，得”-> OK now
|
||||
的 ： Adj/AP ++ "的" ++ NP
|
||||
“地”： Adv ++ “地” ++ V e.g. 飞快(fly fast)地(de)跑(run)
|
||||
“ 得”： V ++ “得” ++ Adv(complement) e.g. 跑(run)得(de)快(fast)
|
||||
|
||||
|
||||
5h Jolene's time, total 25h.
|
||||
|
||||
15/10
|
||||
|
||||
GF/lib/src/chinese/ complete and compilable! Added to darcs by AR.
|
||||
|
||||
Issues from
|
||||
|
||||
Yip-Po Ching and Don Rimmington,
|
||||
Basic Chinese. A Grammar and Workbook,
|
||||
Routledge,
|
||||
London and New York,
|
||||
2009.
|
||||
|
||||
p. 4 the dun-comma in lists DONE
|
||||
|
||||
p. 28 "who is X" vs. "X is who"
|
||||
|
||||
p. 38 er -> liang before a measure word
|
||||
|
||||
p. 41 ordinals require measure words: di yi ge xuesheng -> possessive + ord + class ; no the + ord
|
||||
|
||||
p. 44 how many - duoshao vs. ji -> ji sui(how many years, to kids)
|
||||
|
||||
p. 63 possessive precedes indefinite plural: wo de hen duo pengyou "many of my friends"
|
||||
|
||||
p. 94 degree + adjective + de + noun: hen da de wuzi
|
||||
|
||||
p. 95 no copula in adjectival predication -> use hen
|
||||
|
||||
p. 96 adjectives negated by bu
|
||||
|
||||
p. 97 AB -> AABB reduplication
|
||||
|
||||
p. 104 non-gradable adjectives require shi: zhei tiao yu shi huo de -> keep this as general rule for APs so far
|
||||
|
||||
p. 106 it was Chinese that I studied -> move cleft to the end DONE 16/10
|
||||
|
||||
p. 116 I did it better than you -> to implement
|
||||
|
||||
p. 128 disyllabic place words
|
||||
|
||||
p. 155-158 mapping tenses to aspects
|
||||
|
||||
p. 168 "can": hui/neng
|
||||
|
||||
p. 174 negation and tense
|
||||
|
||||
p. 185 yes/no DONE
|
||||
|
||||
p. 186 alternation questions
|
||||
|
||||
p. 197 "or" haishi/huozhe
|
||||
|
||||
p. 206 "please" in imperative
|
||||
|
||||
p. 207 "let's" zanmen he yi bei ba DONE 16/10
|
||||
|
||||
p. 242 coverbs
|
||||
|
||||
p. 255 disyllabic prepositions
|
||||
|
||||
|
||||
Today's work 6h AR, 3h Jolene. Total 29h.
|
||||
|
||||
Jolene worked 3 weeks for the extended mini Chinese. This means the complete (but not yet perfect) Chinese
|
||||
RGL implementation took 4 person weeks altogether. The remaining immediate fixes shouldn't take a lot of time.
|
||||
However, there is a lot of Chinese that we will leave outside the common RGL abstract syntax.
|
||||
|
||||
|
||||
16/10
|
||||
|
||||
Some more changes in the (almost) complete Chi. It has 546 words:
|
||||
|
||||
0 1 2 3 4 5 6 7 8 9 、 。 一 七 万 丈 三 上 下 不 丑 业 丢 两 个 中 为 么 之 乎 乐 九 乞 书 买 了 争 事 二
|
||||
云 五 些 亮 亲 人 什 今 从 他 以 们 件 会 但 位 何 你 候 做 停 像 儿 光 八 公 六 关 写 冰 冷 净 准 几 则 到
|
||||
刺 前 副 割 劈 加 动 包 匹 医 十 千 单 卖 厂 厌 厚 去 又 友 发 变 只 可 台 右 叶 号 吃 名 后 吐 吗 吧 听 吸
|
||||
吹 呕 呼 和 咬 咱 哪 唉 唱 商 啤 喜 喝 嘴 四 回 因 园 国 圆 土 在 地 场 坏 坐 块 城 堂 堆 塑 处 备 外 多 夜
|
||||
大 天 太 夫 头 奇 套 女 奶 她 好 如 妹 妻 姑 娘 婚 婴 子 字 学 孩 它 宗 定 家 宽 寄 寓 察 对 寻 小 少 尖 尘
|
||||
就 尾 屋 山 工 左 己 已 巴 市 师 希 常 帽 干 平 年 庄 店 座 开 弟 张 往 很 得 心 必 忘 怕 您 想 懂 我 或 战
|
||||
房 所 扇 手 打 扔 把 报 拉 指 挖 挤 推 揉 握 搔 摄 摆 擦 收 敌 教 数 文 料 新 旁 旅 时 明 星 是 晚 暖 更 最
|
||||
月 有 朋 服 望 期 木 本 术 朵 机 杀 村 条 来 杯 板 林 果 枝 架 某 树 校 根 桌 桶 棕 森 棵 椅 橡 欢 歌 止 此
|
||||
死 母 每 比 毛 毯 水 求 汽 沙 没 河 油 法 泳 洗 活 流 浮 海 温 港 游 湖 湿 滑 满 滴 漂 火 灯 灰 炉 烂 烟 烧
|
||||
热 然 爱 父 片 牙 牛 牧 物 狗 猎 猫 王 玩 现 球 理 瓶 生 甲 电 男 画 白 百 的 皮 盏 盐 盒 盖 直 看 眼 着 睛
|
||||
睡 知 短 石 码 破 确 离 种 科 空 窄 窗 站 笑 笔 笨 第 等 答 简 算 箱 粒 红 约 纸 经 绑 结 给 绳 绿 缝 羊 羽
|
||||
翅 翰 老 而 耳 聪 肉 肚 肝 肠 肪 胀 背 胶 胸 能 脂 脏 脖 脚 腿 膀 膝 膨 自 舌 船 艘 艺 花 苹 草 落 蓝 薄 虫
|
||||
虱 虽 蛇 蛋 血 行 衣 表 衫 衬 袜 被 要 规 视 角 言 警 计 讨 语 说 请 读 谁 赢 走 起 趣 跑 距 路 跳 躺 车 转
|
||||
轻 辆 边 过 近 还 这 远 通 道 那 酒 酪 里 重 金 钝 钢 铁 银 长 门 闭 问 间 闻 阳 阵 除 雨 雪 零 雾 非 面 革
|
||||
靴 鞋 音 顶 项 须 颗 题 风 飞 饭 首 马 骨 鱼 鸟 黄 黎 黑 鼻 齿 ！ ， ？
|
||||
|
||||
Also updated the Pinyin version and created a Makefile to convert Chi to Cmn.
|
||||
eng_chi2.txt -- the test set with annotations
|
||||
chinese-log.txt -- a log of the grammar writing process
|
||||
|
||||
There is also a Pinyin version of the grammar, in the subdirectory ./pinyin/.
|
||||
|
||||
The grammar is not yet a part of the "official" demos and documentation. But you can find alpha versions in
|
||||
|
||||
http://tournesol.cs.chalmers.se/~aarne/gf/csynopsis.html -- make sure to select coding UTF8 in your browser
|
||||
|
||||
http://tournesol.cs.chalmers.se:41296/minibar/minibar.html -- select ResourceDemo
|
||||
|
||||
|
||||
|
||||
342
lib/src/chinese/chinese-log.txt
Normal file
342
lib/src/chinese/chinese-log.txt
Normal file
@@ -0,0 +1,342 @@
|
||||
Chinese resource grammar experiment
|
||||
|
||||
(c) Aarne Ranta 2012
|
||||
|
||||
Idea: bootstrap a complete resource grammar by
|
||||
- cloning files from another language (Thai)
|
||||
- extracting lexicon from available sources (Swadesh, HSK list, Google translate)
|
||||
- fixing the errors in syntax and lexicon
|
||||
- testing the grammar in applications
|
||||
|
||||
With next to now knowledge of Chinese, but access to web sources, a grammar book
|
||||
|
||||
Claudia Ross and Jing-heng Sheng Ma,
|
||||
Modern Mandarin Chinese Grammar. A Practical Guide,
|
||||
Routledge,
|
||||
London and New York,
|
||||
2006.
|
||||
|
||||
and an extended mini resource by Jolene Zhuo Lin qiqige
|
||||
and comments on that from Inari Listenmaa. And also Haiyan Qiao's Numeral implementation
|
||||
from 1999.
|
||||
|
||||
The question is how good it will be before the visit to Shanghai in two weeks,
|
||||
and how much work will be needed to fix everything. But so far I feel like the
|
||||
guy in Searle's "Chinese Room", http://en.wikipedia.org/wiki/Chinese_room
|
||||
|
||||
|
||||
5/10/2012
|
||||
|
||||
Clone chinese/*Chi.gf from thai/*Tha.gf by
|
||||
|
||||
runghc lib/src/Clone.hs Tha Chi
|
||||
|
||||
This compiles directly, omitting Lexicon and Structural but producing some Thai words, visible with
|
||||
|
||||
pg -words
|
||||
|
||||
Then port parts of examples/extmini/*Cmn.gf appending them to Lexicon, Structural, Paradigms, Res.
|
||||
Tweak until everything compiles. For simplicity, retain lincat's of Thai.
|
||||
|
||||
Then identify Thai words with 'pg -words', and locate them with 'ma'. Replace them with constants
|
||||
in ResChi, initialized with defaultStr = "". 17 such constants are needed. StringsChi can be eliminated.
|
||||
It was a transient part of Tha anyway.
|
||||
|
||||
Copy examples/numerals/chinese.gf to NumeralChi.gf. Change the "formal" numerals (used mainly for money)
|
||||
to the "normal" ones.
|
||||
|
||||
We now have 47 words (Chinese characters),
|
||||
|
||||
Lang> pg -words
|
||||
0 1 2 3 4 5 6 7 8 9 万 个 他 他们 仟 伍 你 你们 佰 叁 壹 壹万 壹拾
|
||||
壹拾壹 大 女人 她 好奇 小 我 我们 房子 拾 拾万 拾壹 拾壹万 捌 柒 树 棵
|
||||
每 玖 男人 睡 知道 约翰 绿 肆 贰 走 这 那 间 陆 零 非常
|
||||
|
||||
and thousands of linearizable Cl. With Thai word order so far. And all these blank (defaultStr)
|
||||
function words.
|
||||
|
||||
男人 绿 大 非常 每 男人 绿 走 大
|
||||
every green man very bigly is being walking bigly
|
||||
|
||||
约翰 睡
|
||||
John sleeps
|
||||
|
||||
约翰 睡
|
||||
John is sleeping
|
||||
|
||||
他 知道 约翰 走
|
||||
he knows that John would have walked
|
||||
|
||||
每 走
|
||||
everyone walks
|
||||
|
||||
你们 知道 走
|
||||
you know that one will walk
|
||||
|
||||
房子 我们 绿 男人
|
||||
greenest house we is being a man
|
||||
|
||||
这 走 小
|
||||
this is being walking smally
|
||||
|
||||
Ca. 2 hours of work has gone to all this.
|
||||
|
||||
Next step: Chinese Swadesh list, http://en.wiktionary.org/wiki/Appendix:Mandarin_Swadesh_list
|
||||
|
||||
-- № English word POS Pinyin IPA notes Traditional Chinese Simplified Chinese
|
||||
|
||||
> let analyse line = case words line of n:eng:ws | all Data.Char.isDigit n -> eng ++ " " ++ last ws ; _ -> ""
|
||||
> readFile "swadesh.txt" >>= mapM_ (putStrLn . analyse) . lines
|
||||
|
||||
After most of this is done, we have 218 words:
|
||||
|
||||
一些 万 丈夫 个 云 什么 什么时候 他 他们 仟 伍 你 你们 佰 光滑 冰 冷 刺 割 劈开
|
||||
动物 厚 叁 叶 吃 名 吐 听 吵架 吹 呕 呼吸 和 咬 哪里 唱 喝 嗅 嘴 因为 圆 在...里
|
||||
地球 坏 坐 壹 壹万 壹拾 壹拾壹 多 夜晚 大 天 太阳 头 头发 女人 她 好 好奇 如何 如果
|
||||
妻子 孩子 宽 对 小 少 尖 尘土 尾 山 干 年 心脏 怕 想 我 我们 房子 手 打 打猎 扔 拉
|
||||
拾 拾万 拾壹 拾壹万 指甲 挖 挤 捌 推 揉 握 搔 擦 数 新 星 月亮 杀 来 柒 树 树枝 树皮
|
||||
根 森林 棵 死 每 水 水果 沙 河 洗 活 流 浮 海 温暖 游泳 湖 湿 满 火 灰 烂 烟 爱 牙齿
|
||||
狗 玖 玩 男人 白 白天 皮 盐 直 看 眼睛 睡 知道 短 石 种子 窄 站 笑 红 约翰 绑 结冰 绳
|
||||
绿 缝 羽毛 翼 老 耳朵 肆 肉 肚子 肝 肠子 背 胸 脂肪 脏 脖子 脚 腿 膝 膨胀 舌 花 草 落下
|
||||
薄 虫 虱 蛇 蛋 血 角 谁 贰 走 路 躺 转 近 这 这里 那 那里 重 钝 长 间 陆 雨 雪 零 雾
|
||||
非常 风 飞 骨 鱼 鸟 黄 黑 鼻
|
||||
|
||||
We have more than a half left:
|
||||
|
||||
Lang> pg -missing -lang=Chi | ? wc -w
|
||||
329 _tmpi
|
||||
|
||||
But we get more interesting sentences:
|
||||
|
||||
Lang> gr -number=8 PredVP ? ? | l
|
||||
白天 满 冷 多 个 好奇 你们 谁
|
||||
many cold fuller days wonder who you weren't
|
||||
|
||||
胸 黄 红 一些 胸 数
|
||||
some red breast yellowly is counted
|
||||
|
||||
头 握 多 头 看
|
||||
many heads to be held see themselves
|
||||
|
||||
丈夫 那里 对 他 擦
|
||||
his husbands there correctly are wiping them
|
||||
|
||||
约翰 躺
|
||||
John lies
|
||||
|
||||
地球 这里 非常 刺 知道 我们 站
|
||||
earth very here stabbed knows that we wouldn't have stood
|
||||
|
||||
多 那里 那里 想 在...里 她
|
||||
many there there are being thinking in her
|
||||
|
||||
尘土 每 个 那里 握
|
||||
every dust there holds itself
|
||||
|
||||
Ca. 30 minutes for this phase, 2h30 total.
|
||||
|
||||
Now let's take the missing words and try to look them up the in HSK1 word list.
|
||||
Now we have 218 left, less than 200 really missing. But potentially some junk.
|
||||
|
||||
Some more words found in Google translate. Guessing mkVA and other complex subcats.
|
||||
Now we have 126 left, less than 100 really missing. pg -words shows 404 Chinese words:
|
||||
|
||||
…之间 一 一些 万 丈夫 上边 下 不 丢 个 中 为了 为什么 书 买 云 人 什么 什么时候
|
||||
今天 从 从前 他 他们 仟 以后 件 伍 会 但是 你 你们 佰 做 光滑 公寓 写 冰 冰箱 冷
|
||||
刺 前边 割 劈开 动物 医生 卖 厚 原则上 去 叁 发现 变 只 台 右 叶 号 吃 名 吐 吗
|
||||
听 吵架 吸 吹 呕 呼吸 和 和平 咬 哪里 唱 商店 啤酒 喜欢 喝 嗅 嘴 回答 因为 国 国王
|
||||
圆 在 在...里 地板 地毯 地球 坏 坐 块 塑料制成 壹 壹万 壹拾 壹拾壹 多 夜晚 大 大学
|
||||
天 天花板上 太 太阳 头 头发 女人 女王 奶 奶酪 她 好 好奇 如何 如果 妹妹 妻子 姑娘
|
||||
婴儿 存在 学 学校 学生 孩子 它 宗教 宽 寄 对 小 少 尖 尘土 尾 屋顶 山 工业 工厂
|
||||
左 已经 市 希望 帽子 干 干净 年 年轻 开 弟弟 往 心脏 必须 忘 怕 想 愿意 懂 我
|
||||
我们 或者 战 房子 所 手 手套 打 打猎 扔 报 拉 拾 拾万 拾壹 拾壹万 指甲 挖 挤
|
||||
捌 推 揉 握 搔 摄像头 摆 擦 收音机 敌人 教 教堂 数 新 旅行 时间 星 星球 是 是否
|
||||
最 月亮 有 朋友 木 机 杀 村庄 来 柒 树 树枝 树皮 根 桌子 森林 棵 椅子 次 歌 死
|
||||
母亲 每 比 水 水果 汽车 沙 河 油 洗 活 流 浮 海 海港 温暖 游泳 湖 湿 满 漂亮 火
|
||||
火车 灯 灰 炉子 烂 烟 烧 热 爱 父亲 牙齿 牛 牧师 状物 狗 猫 玖 玩 玩儿 现在
|
||||
电视 男人 男孩 画 白 白天 的 的橡胶 的距离 皮 皮革 盐 直 看 眼睛 睡 知道 短
|
||||
石 石头 离 种 种子 科学 窄 窗 站 笑 笔 等 红 约翰 纸 绑 结冰 给 绳 绿 缝 羊
|
||||
羽毛 翼 老 老师 老板 耳朵 肆 肉 肚子 肝 肠子 背 胸 脂肪 脏 脖子 脚 腿 膝 膨胀
|
||||
自己 自行车 舌 船 艺术 花 花园 苹果 草 落下 蓝 薄 虫 虱 虽然 蛇 蛋 血 衣服 表弟
|
||||
衬衫 袜子 被 角 警察 计算机 语法 语言 说 读 谁 贰 赢 走 跑 路 跳 躺 转 近 还是
|
||||
这 这里 远 道理 那 那里 酒 重 重要 金 钝 钢 铁 银 银行 长 门 问 问题 间 陆 除了…
|
||||
以外 雨 雪 零 雾 非常 面包 鞋 音乐 风 飞 飞机 饭店 马 骨 鱼 鸟 黄 黄油 黑 鼻
|
||||
|
||||
|
||||
4h work for a compiling grammar with a decently sized lexicon. The next thing
|
||||
is to fix the worst bugs, in particular in word order.
|
||||
|
||||
|
||||
6/10
|
||||
|
||||
Started a test set for syntax, with 54 sentences testing predication and noun
|
||||
phrase formation. The first run, baseline.txt, uses the Thai-based syntax.
|
||||
|
||||
As the first thing, revised the order of determiners in an NP, and the place of
|
||||
adverbs in VP. Det lincats seem to be too rich, whereas Cl may need to be made
|
||||
discontinuous, to enable the proper place of IAdv. This is correct in Jolene's
|
||||
code.
|
||||
|
||||
3h of work today, total 7h.
|
||||
|
||||
|
||||
7/10
|
||||
|
||||
Ported Jonele's lincat's everywhere, eliminated Thai-style identifiers. Difficult
|
||||
to choose default tenses. But the code is nice to work with.
|
||||
|
||||
4h today, total 11h.
|
||||
|
||||
|
||||
8/10
|
||||
|
||||
Refactored clause building with an overloaded ResChi.mkClause. Added existentials with the verb
|
||||
you_s. Very uncertain on many things, time to call an expert.
|
||||
|
||||
1h today, total 12h.
|
||||
|
||||
Following a suggestion by Thomas Hallgren, divided multicharacter words to as many tokens as
|
||||
there are characters. This means all division into tokens is performed by the parser, which in a
|
||||
sense is optimal. To "remove" spaces in linearization, simply use l -unchars; to "insert" spaces
|
||||
in parsing, use pt -chars | p. All this is done by the oper ResChi.word, which can be changed
|
||||
to change this behaviour.
|
||||
|
||||
|
||||
12/10
|
||||
|
||||
Went through some open questions with Jolene. Then fixed some parameters for Det and Adv and prepared
|
||||
Lexicon and Structural for her inspection and completions.
|
||||
|
||||
5h my time today, 3h Jolene's, total 20h.
|
||||
|
||||
|
||||
14/10
|
||||
|
||||
Lexicon and Structural checked and completed by Jolene. Some open issues found by Jolene, hopefully to be fixed this week:
|
||||
- generics for weather -> Extra
|
||||
- "something" -> Extra (stg for event vs. physical object)
|
||||
- relative clause with complex RP -> postpone; discontinuous?
|
||||
- adjectives for animates -> leave to semantics (cf. Thai)
|
||||
- weather verbs rather VP's -> leave to semantic lexicon
|
||||
- markers 把(marker for direct-object), 被(marker for indirect-object) -> fix for V2A, V3 with ba
|
||||
- positions of the parts of Subj -> fix by introducing particles before or after the subject of the main clause
|
||||
- the three de particles: 的，地，得”-> OK now
|
||||
的 ： Adj/AP ++ "的" ++ NP
|
||||
“地”： Adv ++ “地” ++ V e.g. 飞快(fly fast)地(de)跑(run)
|
||||
“ 得”： V ++ “得” ++ Adv(complement) e.g. 跑(run)得(de)快(fast)
|
||||
|
||||
|
||||
5h Jolene's time, total 25h.
|
||||
|
||||
15/10
|
||||
|
||||
GF/lib/src/chinese/ complete and compilable! Added to darcs by AR.
|
||||
|
||||
Issues from
|
||||
|
||||
Yip-Po Ching and Don Rimmington,
|
||||
Basic Chinese. A Grammar and Workbook,
|
||||
Routledge,
|
||||
London and New York,
|
||||
2009.
|
||||
|
||||
p. 4 the dun-comma in lists DONE
|
||||
|
||||
p. 28 "who is X" vs. "X is who"
|
||||
|
||||
p. 38 er -> liang before a measure word
|
||||
|
||||
p. 41 ordinals require measure words: di yi ge xuesheng -> possessive + ord + class ; no the + ord
|
||||
|
||||
p. 44 how many - duoshao vs. ji -> ji sui(how many years, to kids)
|
||||
|
||||
p. 63 possessive precedes indefinite plural: wo de hen duo pengyou "many of my friends"
|
||||
|
||||
p. 94 degree + adjective + de + noun: hen da de wuzi
|
||||
|
||||
p. 95 no copula in adjectival predication -> use hen
|
||||
|
||||
p. 96 adjectives negated by bu
|
||||
|
||||
p. 97 AB -> AABB reduplication
|
||||
|
||||
p. 104 non-gradable adjectives require shi: zhei tiao yu shi huo de -> keep this as general rule for APs so far
|
||||
|
||||
p. 106 it was Chinese that I studied -> move cleft to the end DONE 16/10
|
||||
|
||||
p. 116 I did it better than you -> to implement
|
||||
|
||||
p. 128 disyllabic place words
|
||||
|
||||
p. 155-158 mapping tenses to aspects
|
||||
|
||||
p. 168 "can": hui/neng
|
||||
|
||||
p. 174 negation and tense
|
||||
|
||||
p. 185 yes/no DONE
|
||||
|
||||
p. 186 alternation questions
|
||||
|
||||
p. 197 "or" haishi/huozhe
|
||||
|
||||
p. 206 "please" in imperative
|
||||
|
||||
p. 207 "let's" zanmen he yi bei ba DONE 16/10
|
||||
|
||||
p. 242 coverbs
|
||||
|
||||
p. 255 disyllabic prepositions
|
||||
|
||||
|
||||
Today's work 6h AR, 3h Jolene. Total 29h.
|
||||
|
||||
Jolene worked 3 weeks for the extended mini Chinese. This means the complete (but not yet perfect) Chinese
|
||||
RGL implementation took 4 person weeks altogether. The remaining immediate fixes shouldn't take a lot of time.
|
||||
However, there is a lot of Chinese that we will leave outside the common RGL abstract syntax.
|
||||
|
||||
|
||||
16/10
|
||||
|
||||
Some more changes in the (almost) complete Chi. It has 546 words:
|
||||
|
||||
0 1 2 3 4 5 6 7 8 9 、 。 一 七 万 丈 三 上 下 不 丑 业 丢 两 个 中 为 么 之 乎 乐 九 乞 书 买 了 争 事 二
|
||||
云 五 些 亮 亲 人 什 今 从 他 以 们 件 会 但 位 何 你 候 做 停 像 儿 光 八 公 六 关 写 冰 冷 净 准 几 则 到
|
||||
刺 前 副 割 劈 加 动 包 匹 医 十 千 单 卖 厂 厌 厚 去 又 友 发 变 只 可 台 右 叶 号 吃 名 后 吐 吗 吧 听 吸
|
||||
吹 呕 呼 和 咬 咱 哪 唉 唱 商 啤 喜 喝 嘴 四 回 因 园 国 圆 土 在 地 场 坏 坐 块 城 堂 堆 塑 处 备 外 多 夜
|
||||
大 天 太 夫 头 奇 套 女 奶 她 好 如 妹 妻 姑 娘 婚 婴 子 字 学 孩 它 宗 定 家 宽 寄 寓 察 对 寻 小 少 尖 尘
|
||||
就 尾 屋 山 工 左 己 已 巴 市 师 希 常 帽 干 平 年 庄 店 座 开 弟 张 往 很 得 心 必 忘 怕 您 想 懂 我 或 战
|
||||
房 所 扇 手 打 扔 把 报 拉 指 挖 挤 推 揉 握 搔 摄 摆 擦 收 敌 教 数 文 料 新 旁 旅 时 明 星 是 晚 暖 更 最
|
||||
月 有 朋 服 望 期 木 本 术 朵 机 杀 村 条 来 杯 板 林 果 枝 架 某 树 校 根 桌 桶 棕 森 棵 椅 橡 欢 歌 止 此
|
||||
死 母 每 比 毛 毯 水 求 汽 沙 没 河 油 法 泳 洗 活 流 浮 海 温 港 游 湖 湿 滑 满 滴 漂 火 灯 灰 炉 烂 烟 烧
|
||||
热 然 爱 父 片 牙 牛 牧 物 狗 猎 猫 王 玩 现 球 理 瓶 生 甲 电 男 画 白 百 的 皮 盏 盐 盒 盖 直 看 眼 着 睛
|
||||
睡 知 短 石 码 破 确 离 种 科 空 窄 窗 站 笑 笔 笨 第 等 答 简 算 箱 粒 红 约 纸 经 绑 结 给 绳 绿 缝 羊 羽
|
||||
翅 翰 老 而 耳 聪 肉 肚 肝 肠 肪 胀 背 胶 胸 能 脂 脏 脖 脚 腿 膀 膝 膨 自 舌 船 艘 艺 花 苹 草 落 蓝 薄 虫
|
||||
虱 虽 蛇 蛋 血 行 衣 表 衫 衬 袜 被 要 规 视 角 言 警 计 讨 语 说 请 读 谁 赢 走 起 趣 跑 距 路 跳 躺 车 转
|
||||
轻 辆 边 过 近 还 这 远 通 道 那 酒 酪 里 重 金 钝 钢 铁 银 长 门 闭 问 间 闻 阳 阵 除 雨 雪 零 雾 非 面 革
|
||||
靴 鞋 音 顶 项 须 颗 题 风 飞 饭 首 马 骨 鱼 鸟 黄 黎 黑 鼻 齿 ！ ， ？
|
||||
|
||||
Also updated the Pinyin version and created a Makefile to convert Chi to Cmn.
|
||||
|
||||
|
||||
19/10
|
||||
|
||||
Evaluation results for Synopsis terms from Jolene, file eng_chi2.txt. Summary:
|
||||
|
||||
GOOD 295 68%
|
||||
SOSO 66 15%
|
||||
BAD 76 17%
|
||||
|
||||
The annotation guidelines were:
|
||||
|
||||
GOOD -- grammatically correct rendering of the English meaning
|
||||
SOSO -- grammatically correct in some conditions, but means something different, or is strange with these words
|
||||
BAD -- grammatically incorrect
|
||||
|
||||
Some statistics on the BAD's:
|
||||
|
||||
10 "constant not found" errors, for structural words.
|
||||
5 in Numerals - maybe result from my bad conversion of Qiao's code?
|
||||
|
||||
|
||||
|
||||
1753
lib/src/chinese/eng_chi2.txt
Normal file
1753
lib/src/chinese/eng_chi2.txt
Normal file
File diff suppressed because it is too large
Load Diff
Reference in New Issue
Block a user