Could this type of language be used in the future to ease natural language processing pains?
No, on two counts:
It's hardly a breakthrough in natural language processing to shift load onto the human by making them learn a new language. What do you think "typing" is but a specialized sign language? Making them learn a new language defeats the whole purpose and makes for a rather hollow victory.
While "word rate" varies somewhat from culture to culture, "information rate" is basically a constant. To express "The little boy was hit by a blue ball and started to cry, but his mother cheered him up with some cookies." will take about the same amount of time in spoken langauge in all languages (meant for face-to-face interaction).
(It's actually somewhat surprising that there's as much varience as there is in the length of the written version of that sentence; you can see in many languages that speaking has been more importent then writing. I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages. Thank the Anglo-Saxons.)
Creating an acoustically simpler language will necessarily mean that artificial language will be slower to communicate with. (If you could communicate at the same rate as English, then by pretty much by definition it would as complex.) Again, "reducing" the problem like this isn't so impressive and doesn't really solve the problem.
And that's assuming what you really meant was "speech recognition pains". The real problem with "natural language recognition" is the stupifyingly complex sentences we utter, with their amazing context-sensitivity and ambiguities. NLP isn't a solved problem even on plain text which removes the vast majority of acoustical ambiguities that speech recognition has to deal with. (You still have problems like "ram" (verb, noun), but that's part of NLP.)
Basically, this is not useful for human-computer interaction. Limited forms of it have been useful in the other direction, though, but I don't know how the sounds mapped to information. AFAIK jet-fighter cockpits use acoustic signals, but they aren't used to convey digital information like words, they convey analog information like distances or speeds.
I think the assumption you are making is that NLP and SR have to convey the same amount of information to a computer as speach does to a human being.
In programming you have literal values for things like text strings that don't modify the semantics of your program. The rest of the program, following the syntax of some language (like C), is all that the computer needs to understand what you want it to do. Notice that the syntax for your common programming language are quite terse (needfully so).
I think the assumption you are making is that NLP and SR have to convey the same amount of information to a computer as speach does to a human being.
Well, I wouldn't call it an "assumption" so much as the "definition of NLP". The whole point of NLP is to get as much out of the text as humans do. If you're happy about constraining the language in advance, it's a lot easier... but then it's not natural language processing, it's unnatural language processing.
I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages.
Doesn't that already happen?
First of all, English is the dominant language in worldwide communication (if somebody now wants to remind me of the Chinese dialects, Hindi, and Spanish: I said worldwide communication, not most prominent mother languages). This will automatically affect the other languages
This reminds me of my favourite English-language quote...
"The problem with defending the purity of the English language is that English is about as pure as a cribhouse whore. We don't just borrow words; on occasion, English has pursued other languages down alleyways to beat them unconscious and rifle their pockets for new vocabulary."
I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages. Thank the Anglo-Saxons.
You mean thank the Angles and the Saxons (two different groups). And what's with the prejudice against the Jutes?
While "word rate" varies somewhat from culture to culture, "information rate" is basically a constant. To express "The little boy was hit by a blue ball and started to cry, but his mother cheered him up with some cookies." will take about the same amount of time in spoken langauge in all languages (meant for face-to-face interaction).
Is that true for Hawaiian, too? What with their reduced phoneme bandwidth and all? Or do they spit their phonemes out faster?
Listen to a Japanese speaker, such as a voice actor in subtitled anime. Japanese has a reduced nominal phoneme inventory, fewer than that of Spanish but a bit more than that of 'Nesian languages. Japanese speakers do spit out phonemes quickly, so quickly in fact that speakers often elide unaccented 'i' and 'u' to palatalization or rounding of the previous consonant.
Worse yet, listen to a Toki Pona speaker. Toki Pona's minimalist phoneme inventory compares to
So you are kindof saying that some languages do have lower (idea) bandwidth? I understand that most languages have similar basic capacities, but I always thought there'd have to be stand-outs.
(It's actually somewhat surprising that there's as much varience as there is in the length of the written version of that sentence; you can see in many languages that speaking has been more importent then writing. I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages. Thank the Anglo-Saxons.)
Are you sure about this? I actually lost an argument with a native G
I checked up and there's been extensive brain research on this language. It turns out that the whistling follows key points in the Spanish sounds, and that the brain reacts to them in the same way as when listening to those same words in Spanish.
In other words, all signal information other than those key points is ignored by the brain.
This fact can be used in speech recognition, as it tells us we can dispense with most of the signal, and only look at the key points. This will allow computers to handle a
It's hardly a breakthrough in natural language processing to shift load onto the human by making them learn a new language. What do you think "typing" is but a specialized sign language? Making them learn a new language defeats the whole purpose and makes for a rather hollow victory.
Ecept that I can type faster then I can talk. Much like reading, when I type I don't have to 'sound it out' so I'm willing to bet that, also like reading, the control processes are different.
Welcome to/. - I can't be the only introvert. I do type, depending upon context 100-120 WPM. I rarely talk that fast. Nor do i talk as much as I type. Nor do I talk as often. YMMV
"I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages"
While letter-based entry is easy for computers to be designed to do, symbol based output is much more efficient than letter based words. It's a lot easier to write fast car in Japanese -- 2 characters (one for speed, one for car) vs. 7 in English + space or hyphenation.
(It's actually somewhat surprising that there's as much varience as there is in the length of the written version of that sentence; you can see in many languages that speaking has been more importent then writing. I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages. Thank the Anglo-Saxons.)
Actually you should thank the Normans probably, and the old French th
This was also accellerated by Anglo Saxon not being used as a formal and written language for several hundred years as pretty much everything being used for formality was either in Latin (in the Church) or Norman French (almost everywhere else).
So the simplifications (note that this is not the same as decline) that are always made in spoken language tended to be incorporated within the language as a whole, while anyone interested in grammar was looking the other way.
Figuring out if "ram" is a verb or noun sounds easy compaired to the next problem: Is the noun "ram" an animal, a piece of computer hardware or maybe a car.
I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages.
No wA! ppl wl nvr tlk lk dat! w@ r U, %-)?
...there can be no public or private virtue unless the foundation of action is
the practice of truth.
- George Jacob Holyoake
Processing power is a constant (Score:5, Insightful)
No, on two counts:
- It's hardly a breakthrough in natural language processing to shift load onto the human by making them learn a new language. What do you think "typing" is but a specialized sign language? Making them learn a new language defeats the whole purpose and makes for a rather hollow victory.
- While "word rate" varies somewhat from culture to culture, "information rate" is basically a constant. To express "The little boy was hit by a blue ball and started to cry, but his mother cheered him up with some cookies." will take about the same amount of time in spoken langauge in all languages (meant for face-to-face interaction).
And that's assuming what you really meant was "speech recognition pains". The real problem with "natural language recognition" is the stupifyingly complex sentences we utter, with their amazing context-sensitivity and ambiguities. NLP isn't a solved problem even on plain text which removes the vast majority of acoustical ambiguities that speech recognition has to deal with. (You still have problems like "ram" (verb, noun), but that's part of NLP.)(It's actually somewhat surprising that there's as much varience as there is in the length of the written version of that sentence; you can see in many languages that speaking has been more importent then writing. I suspect over the next hundred years some of the more verbose letter-based written languages will start condensing down to be more like English, which is one of the more compact letter-based languages. Thank the Anglo-Saxons.)
Creating an acoustically simpler language will necessarily mean that artificial language will be slower to communicate with. (If you could communicate at the same rate as English, then by pretty much by definition it would as complex.) Again, "reducing" the problem like this isn't so impressive and doesn't really solve the problem.
Basically, this is not useful for human-computer interaction. Limited forms of it have been useful in the other direction, though, but I don't know how the sounds mapped to information. AFAIK jet-fighter cockpits use acoustic signals, but they aren't used to convey digital information like words, they convey analog information like distances or speeds.
Re:Processing power is a constant (Score:1)
In programming you have literal values for things like text strings that don't modify the semantics of your program. The rest of the program, following the syntax of some language (like C), is all that the computer needs to understand what you want it to do. Notice that the syntax for your common programming language are quite terse (needfully so).
If your
Re:Processing power is a constant (Score:2)
Well, I wouldn't call it an "assumption" so much as the "definition of NLP". The whole point of NLP is to get as much out of the text as humans do. If you're happy about constraining the language in advance, it's a lot easier... but then it's not natural language processing, it's unnatural language processing.
There's nothing "wrong" with that, of course, bu
Re:Processing power is a constant (Score:1)
Doesn't that already happen?
First of all, English is the dominant language in worldwide communication (if somebody now wants to remind me of the Chinese dialects, Hindi, and Spanish: I said worldwide communication, not most prominent mother languages). This will automatically affect the other languages
Re:Processing power is a constant (Score:1)
"The problem with defending the purity of the English language is that
English is about as pure as a cribhouse whore. We don't just borrow words;
on occasion, English has pursued other languages down alleyways to beat
them unconscious and rifle their pockets for new vocabulary."
- James D. Nicoll
Which does seem to have some basis in truth...
Re:Processing power is a constant (Score:2)
You mean thank the Angles and the Saxons (two different groups). And what's with the prejudice against the Jutes?
Re:Processing power is a constant (Score:2)
Is that true for Hawaiian, too? What with their reduced phoneme bandwidth and all? Or do they spit their phonemes out faster?
Elision is one key (Score:1)
Or do they spit their phonemes out faster?
Listen to a Japanese speaker, such as a voice actor in subtitled anime. Japanese has a reduced nominal phoneme inventory, fewer than that of Spanish but a bit more than that of 'Nesian languages. Japanese speakers do spit out phonemes quickly, so quickly in fact that speakers often elide unaccented 'i' and 'u' to palatalization or rounding of the previous consonant.
Worse yet, listen to a Toki Pona speaker. Toki Pona's minimalist phoneme inventory compares to
Re:Elision is one key (Score:2)
Re:Processing power is a constant (Score:2)
Are you sure about this? I actually lost an argument with a native G
Perhaps the wrong basis? (Score:2)
In other words, all signal information other than those key points is ignored by the brain.
This fact can be used in speech recognition, as it tells us we can dispense with most of the signal, and only look at the key points. This will allow computers to handle a
Re:Processing power is a constant (Score:2)
Re:Speaking = 200 WPM (Score:2)
Why go with redundancy? (Score:2)
While letter-based entry is easy for computers to be designed to do, symbol based output is much more efficient than letter based words. It's a lot easier to write fast car in Japanese -- 2 characters (one for speed, one for car) vs. 7 in English + space or hyphenation.
Before you whine that learning a few
Thank the Normans (Score:2)
Actually you should thank the Normans probably, and the old French th
Re:Thank the Normans (Score:2)
So the simplifications (note that this is not the same as decline) that are always made in spoken language tended to be incorporated within the language as a whole, while anyone interested in grammar was looking the other way.
However, that's not to say that when En
Re:Processing power is a constant (Score:1)
Re:Processing power is a constant (Score:1)
No wA! ppl wl nvr tlk lk dat! w@ r U, %-)?