Since, presumably, the 140 character limit derives from SMS message limits, it might be interesting to look at what the message lengths are for these (note that this is GSM specific and I don't know the CDMA equivalents):
SMS payload is 140 octets. That means that you can store 140 8bit characters, eg Latin1 encoding, in a text message.
However, the GSM default encoding is 7bit packed. Each character is 7bit and they are packed together into the 140 octets, this leaves a maximum size of 160 characters per message.
For international characters, GSM uses USC2, which is a 16bit encoding, limiting messages to 70 characters.
You can also send concatenated messages, where multiple SMS messages make up the actual message text. In this case, you must subtract (from memory, so may be off) 4 octets from the limits, for the user data header. The user data header consists of a length and one or more information elements. An information element consists of 3 octets: an information element identifier (in this case, 0, for concatenated message) and some parameters (I'm not sure if this is specific to the IEI? I could look it up, but I'm too lazy.. for concatenated messages, the next octet is number of messages and the final one is the current message index).
Hopefully someone found this information interesting ;-)
Minor correction - information elements consist of an information element identifier, a length and length number of octets for a total of "length + 2" octets per information element. In the case of concatenated messages, IE identifier is 0 and length is 3 and the 3 octets of data are id, part and number of parts, in that order.
It's an interesting experiment, but I don't trust the conclusion. Automatic translation still tends to be fairly literal, rather than idiomatic. I have a suspicion that this tends to increase the average message length. An interesting experiment would be to translate English tweets into those various target languages, and compare the two. If my hunch is right, the text will get bigger in both directions. The ratio between the increases may be a good measure of the information density of idiomatic expression.
You're right, it's absolutely unscientific and the results are of dubious statistical significance. Going through machine translation makes the whole thing pointless anyway.
But it was fun to write, which I suppose was the actual point of it. One of the joys of being able to program computers is that when you want to have such a question answered, you can a short bit of code to get the answer.
SMS payload is 140 octets. That means that you can store 140 8bit characters, eg Latin1 encoding, in a text message.
However, the GSM default encoding is 7bit packed. Each character is 7bit and they are packed together into the 140 octets, this leaves a maximum size of 160 characters per message.
For international characters, GSM uses USC2, which is a 16bit encoding, limiting messages to 70 characters.
You can also send concatenated messages, where multiple SMS messages make up the actual message text. In this case, you must subtract (from memory, so may be off) 4 octets from the limits, for the user data header. The user data header consists of a length and one or more information elements. An information element consists of 3 octets: an information element identifier (in this case, 0, for concatenated message) and some parameters (I'm not sure if this is specific to the IEI? I could look it up, but I'm too lazy.. for concatenated messages, the next octet is number of messages and the final one is the current message index).
Hopefully someone found this information interesting ;-)