4 ms·
Calling asl "signed english" is usually a good indication that the writer doesn't know much about the subject. Then they go on to say: > "It turns out that th
by cbr 17y ago
Calling asl "signed english" is usually a good indication that the writer doesn't know much about the subject. Then they go on to say:
> "It turns out that the information content of handshapes is on average just 0.5 bits per handshape less than the theoretical maximum. By contrast, the information content per phoneme in spoken English is some 3 bits lower than the maximum."
Talking about the information content of speech symbols is likely to be entirely bunk, but I'm going to go read the full article and try to find out whether the summary is bad or if the research is really this confused.
- cbr 17y agoJust read the full article. It's no good. Analyzing ASL in terms of hand shapes only, on the grounds that they're the only feature that all linguists looking at ASL seem to agree on is really problematic. Computation of entropy is heavily dependent on the alphabet. If you say english has an alphabet at the syllable level, you'll get a very different amount of entropy than if you say the alphabet is phoneme level. If you go a level below phonemes and look at distinctive features (voiced/unvoiced, aspriated/unaspriated, front/back, high/low, ...) you'll get yet another entropy number altogether. These numbers will be wildly different from each other. Entropy calculations between languages must use comparably derived alphabets which cover the whole communication channel, or they will be meaningless. Argh. Junky article.
- ars 17y agoHow would you calculate the theoretical maximum for a handsign? Or for spoken language for that matter.