search

VictorTaelin / Bitspeak

MIT License
81 stars 4 forks source link

The diphthongs in the 8-bit variant are very difficult to distinguish by ear #4

Open tsbockman opened 4 years ago

tsbockman commented 4 years ago

The diphthongs ai, ei, aw, ow, and even ui all have pronunciations that overlap with those of the core single vowels a, e, i, o, u in some way. Even when they are distinguished, the difference is often subtle enough that people aren't very good at consistently pronouncing and hearing it.

I suggest replacing t in the 8-bit starting consonant table with h and replacing the diphthongs ai, ei, ui, aw, ow in the vowel table with with at, et, it, ot, and ut.

That just leaves oi, which is the easiest of the diphthongs in the table to distinguish from the single vowels.

VictorTaelin commented 4 years ago

Why a t exactly? Being a Portuguese speaker I'm not sure what the sound of that would be, can I see an example? Perhaps an s or r would be more effective.

tsbockman commented 4 years ago

Using s to end some syllables and h to start others creates a written ambiguity because you already used sh as a separate symbol: pasha could be pa-sha or pas-ha. There would be no way to tell from how it's spelled which bit pattern is intended.

Using r would work in written form. However, when h appears directly after another consonant, it often becomes silent, or just slightly alters the next vowel. For example, there is no h sound in rhodes; it sounds like rodes. This same issue occurs in combination with many other consonants in English.

However, t specifically does not cause this problem: patha would be pronounced as either path-a or pat-ha, both of which can be easily distinguished from all other sounds currently in Bitspeak. (Just like sh and ch, th in English normally combine with a unique sound: either a voiceless or voiced dental fricative.)

Your other ending consonant, n, is also OK: I think that panha would always be pronounced pan-ha.

tsbockman commented 4 years ago

I was thinking about the problem of a silent h in rh a bit more, and realized that it mostly occurs at the beginning of words. Since Bitspeak would only ever generate rh in the middle of words, I think people would pronounce the h.

So, while I still think t is better, r is fine. But, definitely don't use s.

VictorTaelin commented 4 years ago

TBH in general I think the 8-bit version is just worse and full of ambiguities. I'm not sure it can be saved. The 6-bit one is the most accurate I can think of, and even it still has some minor issues.

tsbockman commented 4 years ago

You're right that the 6-bit one is better. And, I thought of a simple way to extend it so that it can do everything the 8-bit version does, without the problems:

Add h (or w) as a consonant that represents nothing, and o as a vowel that represents nothing. Examples:

Hexadecimal 3b is 0011 1011 in binary, which becomes dihe in Bitspeak. Hexadecimal 9127 is 1001 0001 0010 0111 in binary, which becomes vakijo in Bitspeak.

In this way, any even number of bits can be expressed precisely in Bitspeak, while retaining all the advantages of your current 6-bit scheme. The existing 8-bit scheme can simply be removed, as it is unnecessary.

johnchandlerburnham commented 4 years ago

I used a terminal n in Syllabytes for 8-bit https://github.com/johnchandlerburnham/syllabytes, but this requires a ' character to disambiguate from an initial n