I'd say word-based dictionary attack is more an alphabetic language practice and not that effective on a character-based langauge like Chinese.
Most Chinese characters can be seen as "words" on their own and Chinese people often think in units of characters instead of words. For example, they use 字典 (literally "character book", dictionary of characters) instead of 词典 ("word book", dictionary of words). So it's natural for a Chinese person to come up with character sequences that make sense but does not show up in any word list (it's like the Chinese version of
For example, the example in Tom's answer,
北极熊, is composed of 北(north), 极(extrme, understood as pole when following 北) and 熊 (bear). Someone could easily come up with
南极熊 (south pole bear) as their password and this might not appear in any word list as it's not a real thing. Simiarly, most people's names does not make any sense as words and they're very likely to use them in passwords.
A more thorough approach would be to come up with all possible character pronunciations in Pinyin and use those in your dictionary. This will be like creating an alphabet for Chinese. This could explain why the "word list" in some sources are much smaller.
You can of course add multi-character word to speed up the process.