I'm looking to test some Morphological Analysis tools, which split sentences into tokens and provide part of speech and pronunciation information (specifically, providing kana when given kanji).

I'm after some sentences which automated tools would have trouble with. Examples of what I'm after are sentences which are difficult to determine the word boundaries, pronunciation or part of speech. Preferably, examples would be a more correct interpretation, and be something that could be seen in actual use.

The best I have found so far is "私はあくまで執事ですから". In a similar vain, is "にわにはにわとりがいる (庭には二羽鶏が居る)" which only applies to spoken use. An english example might be something like "Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo", which is difficult to determine the part of speech each word is.

  • 3
    in meiji era, the kana notation of 庭 was ニハ. so the above sentence was ニハニハニハニハトリガヰル. another variation 裏庭には二羽庭には二羽鶏がいる is also well known. – jovanni Oct 14 '13 at 19:00
  • 2
    In particular, I encountered one type of ambiguity while learning Japanese. The は particle marks a topic, so it is perfectly fine to say 僕は鰻だ (I'm an eel) when a waiter asks what you are having, because it may also be interpreted as (About me, (my order) is eel). – user4060 Oct 15 '13 at 20:39
  • sf.airnet.ne.jp/ts/language/unagi.html claims the said sentence is a case of abbreviation, not topic-marking. – ithisa Oct 18 '13 at 13:45

I've found several webpages online which talk about the sentence 黒い髪のきれいな女の子 as a famous example of an ambiguous sentence. (For example, this blog post).

Here are a couple ways in which this sentence could be considered ambiguous:

  • 女の子 could mean a female child or a woman's child.
  • 黒い could modifies 髪, meaning black hair, or it could modify (髪のきれいな)女, meaning black woman with beautiful hair, or it could modify (髪のきれいな女の)子, meaning black child (of the woman with beautiful hair).

Some of the ambiguities, I think, come from reading the sentence in unusual ways.

  • I think in "default" parsing (i.e. my non-native unreliable Japanese intuition), の binds the most loosely and な bind stronger, and い binds even stronger, so I would instantly read it as [[[黒い[髪の][[きれいな][女の子]]]]] – ithisa Oct 15 '13 at 1:43

I'm looking to test some Morphological Analysis tools, which split sentences into tokens and provide part of speech and pronunciation information.

One of the beauties of Japanese is that there is no need for "pronunciation information". No diphthongs, "silent e", etc. Knowing the 振り仮名 is enough.

Anyway, not sure of your purpose, but how about this old classic:

another test
This will tokenize easily, but there is zero chance it will resolve 「触れない」 to the correct kana.

「触れない」-->「ふれない」」 -->「ふれる」 -->「触れる」

  • By pronunciation information, I simply meant it provides kana when given kanji. That example is only difficult to parse in spoken Japanese. I'm after written Japanese examples. – Spacehamster Oct 4 '13 at 23:35

1. ここではきものを脱いでください
a. ここで[履物]{はきもの}を脱いでください。(here, please put off your shoes)
b. ここでは[着物]{きもの}を脱いでください。( put off your clothes here )

2. このはしを渡ってはいけません
a. don't walk through this bridge.
b. keep center when walking through this.


I think that you are looking for written expressions in Japanese which are ambiguous at some level.

Of course, an easy way to get ambiguity in written text is by using homographs, which is often resolved by considering the meanings:

書の[大家]{たいか} authority of caligraphy
アパートの[大家]{おおや} the owner of apartment

彼女は京都に[行]{い}った。 She went to Kyoto.
彼女は消防設備の点検作業を[行]{おこな}った。 She performed the inspection of fire protection equipments.

([行]{おこな}った is often written as 行なった exactly to avoid this ambiguity.)

In Japanese, the word boundaries and the parts of speech are often clear from the form of the words when a sentence is written with kanji. Therefore it is difficult to get ambiguity in word boundaries and parts of speech. Here is a somewhat close example I managed to make up:


It is incorrect to parse 立場上手 as a compound word consisting of 立場 (たちば) and 上手 (the common spelling of homographs うわて, かみて, and じょうず). The correct way to parse 立場上手 is an adverb 立場上 (which is made by adding suffix -上 to noun 立場) which modifies the rest of the sentence, and a noun 手:

この件に関して彼は[立場上]{たちばじょう}[手]{て}を出すことができない。 He is not in a position to do anything about this matter.

  • There is an internet meme “○○がこの先生きのこるには”, which is considered to be funny because of its garden-path property. Similarly people sometimes talk about how newspaper heading “○○の信頼損なう” looks as if it uses Twitter slang なう. – Tsuyoshi Ito Oct 18 '13 at 16:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.