Hacker News Reader

Comment by philliphaydon | Hacker News Reader

I'm unsure how this works, because this technique wont teach you what the English equiv is when 2 (or more) words are next to each other are.

For example, COW + MEAT = BEEF.

牛肉

牛 = Cow

肉 = Meat

This is a very simple example but there are more complicated words such as Divorce.

In Chinese this is 離婚 which is 離 (from or without) and 婚 (marriage) but it's hard to learn that this is actually Divorce in English.

Edit: Wife is Taiwanese and I'm trying to learn Chinese.

View on HN

Both your examples seem to be a problem for someone learning english (i.e. there's a specific word for cow meat), but not a problem for someone learning Chinese. Perhaps I'm missing your point, but if I come across a word that translates to "cow meat", I also know that as "beef" in english, but it's perfectly understandable to me as "cow meat".

But going the other way, for someone learning english though, this is a problem. They may very well understand and know "cow meat", but have no idea about the word "beef" if they have not encountered it before.

Sure I know cow meat is beef. But what I find hard is knowing when 2/3/4 characters together mean a specific English word. So when trying to read something I find it hard to figure out when I should and shouldn’t combine characters. I’m struggling a bit to learn. Picking up speaking better for me.

Exactly.

Just like 对不起人 translates to “Canadians”

;)

this was good :-)

It isn't picking out just the Hanzi - it is picking out words! If you look at the second picture used it detects the top 3500 words from a .YAML file (and expects you to add words that are not detected in the top 3500 words).

For example, 離婚 is in the file.

    - trad: 離婚
      simp: 离婚
      pinyin: lí hūn
      defs:
      - to divorce
      - divorced from (one's spouse)
      example_sentences:
      - trad: 我想她會和他離婚。
        simp: 我想她会和他离婚。
        pinyin: wǒ xiǎng tā huì hé tā líhūn 。
        eng: I think she will divorce him.

牛肉 is not in the file but is such a simple example. 牛奶 made the list though! So I guess milk is more common than beef.

Yeah, the list isn't perfect (for example, 羊肉 is listed even though 牛肉 isn't). 牛肉 isn't common enough to make the list, but for some reason the Chinese government decided 羊肉 should be one of the first 300 words you learn (it's in HSK 2), so it ended up high on my list because my list partially pulls from HSK, in addition to another list called SUBTLEX-CH.

In this case, I think actually neither should be on the list, and you should instead learn 牛，羊，and 肉 separately and infer the meaning of 牛肉 and 羊肉 from those. I'll change it.

How do you learn to infer the meaning of the 2 put together? Is there a method or is it something you just learn over time?

I built my own dictionary with 114,326 words for Pingtype. I based it on the word spacing from the romanised Bible and Apple's option-right-arrow method, and I've made over 5000 edits.

In my opinion, reading traditional characters with spaces is much easier than simplified characters without spaces.

I don't study Chinese, but I do study Japanese which shares the issue of combination words being multiple kanji (not a big surprise since kanji come from hanzi after all...)

It's something you learn over time for the ones that are more obvious at least. Not all combination words are straight forward, but many basic words are straightforward. "羊肉" is literally "lamb meat" and if you're told 牛 means "cow" you can probably guess "牛肉" means "cow meat". And if 豚 means "pig", what do you think "豚肉" means? And 魚 means "fish" so "魚肉" means? They aren't always so perfectly straight forward, but I bet you can guess what this means too: 消防士 "extinguish (a fire), defend, gentlemen."

This type of meaning inference doesn't always work, though it will for the vast majority of what many would consider "basic vocabulary".

This is actually what makes learning Chinese much easier. Very often you can guess the correct word even when you don't know it. (Such as pingguojiu for cider, a real life example)

Looks like it segments the dictionary and then looks it up. It could still be a problem if it oversegments, but if you look at the screenshots, most "words" are two characters.

The segmentation is based on a library called jieba. It seems to work pretty well from my experience. https://github.com/fxsjy/jieba

I do notice that some generated words are bogus (compounds of other words or redundant in some way). I have a file that lists redundant words, and when I notice these words I add them to the file so the tool won't generate them again. It also lists the words that they are duplicates of, so that those words can be upranked. This is the file if you're curious: https://github.com/kerrickstaley/Chinese-Vocab-List/blob/mas...

Sure, there are words that are relatively opaque based on their components, but 離婚 isn't one of those words. Look at the 6 definitions given for 離 in the ABC dictionary:

1. (verb) leave; part from; be away from -- 我的夫人離我而去 My wife left me.

2. (verb) separate

3. (verb) defy; go against

4. (preposition) distant / apart from -- 北京離上海有多远？ How far is Beijing from Shanghai?

5. (noun) name of one of the 8 Trigrams

6. (noun) name of one of the 64 Hexagrams

The sense of 離 is separation; it makes perfect compositional sense in 離婚. A divorce involves two people starting together and going apart.

From the description, it sounds like it works with words, not characters, so 牛肉 would be one multi-character word. The screenshot even shows an example: 燃烧

That said, I wonder if it also creates cards for the individual characters when they make sense. I know that for me, it's often easier to learn a composite word if I know the individual characters, and it helps you figure out other likely words (eg: cow+ meat = beef, so pig+meat = pork).

Some single-character words are included in the vocabulary list if they're often used as stand-alone words. Others that are rarely used alone or whose meaning can be inferred from "compound" words (e.g. 边, which occurs in 旁边，右边，左边，etc) are omitted. It's all based on my best judgement (which is probably misguided sometimes). A lot of manual tweaking went into generating the vocab list. All this deduplication is encoded in this file if you're curious: https://github.com/kerrickstaley/Chinese-Vocab-List/blob/mas...

Also, in order to generate a flashcard for such a word, it would have to appear in a standalone context, e.g. 我没吃过牛肉 would not generate flashcards for 牛 or 肉, but 牛不吃肉would.

Not to be pedantic, but I think 離 in this case is abbreviated for 離開 (to leave). So divorce means to leave a marriage...

NER might help here.

Reply on news.ycombinator.com