7 ms·

Attempto Controlled English

100 points by sublinear 3y ago

dang 3y ago
Related: Attempto Controlled English - https://news.ycombinator.com/item?id=20950126 https://news.ycombinator.com/item?id=20950126 - Sept 2019 (60 comments) AceWiki:an open-source wiki, that can understand english(almost) - https://news.ycombinator.com/item?id=1563384 https://news.ycombinator.com/item?id=1563384 - July 2010 (1 comment)
oezi 3y ago
In the age of LLMs this seems like a waste of time to even think about constraining and formalizing languages. Why should human users restrict themselves in their expression if LLM systems can easily point out any remaining ambiguities?
- ggm 3y ago
  What they seem to be doing at the moment is riffing on the ambiguities. If they point them out, then the outcome will be .. less ambiguous language which will tend to be like ACE. Gimme the ball is idiomatic. How do you know it means Give me, and not Furble the ball? We're taught Give me, we say Gimme. Both have a role.
  - HPsquared 3y ago
    It could also mean "please give me the management role overseeing the event known as the ball".
- tgv 3y ago
  I'm interested to see an LLM do that. Technically, nearly every sentence is ambiguous, but that clearly doesn't bother us. An automated system that can point out sentences humans are likely to interpret in multiple ways would be interesting. However, no such thing seems to exist.
  - fallingmeat 3y ago
    I recently made an LLM-based tool to identify inconsistencies with the INCOSE guide to writing requirements (identifying vague language, for example). Also have had some success translating those outputs to more structured language like LTL. I think it is doable
  - oezi 3y ago
    I am curious why you think this. A basic prompt of pointing out ambiguities should get you quite far. I am pretty sure LLMs would be great to turn regular English into ACE including resolving detected ambiguities.
  - euroderf 3y ago
    When working as a tech writer, I was mildly obsessed with this. Nothing drives users away from your documentation like a lack of clarity that signals that (1) you're on your own now, pal, and (2) the writer had zero clue of the potential for chaos in interpreting what he is writing.
- stareatgoats 3y ago
  You make a valid point, LLMs intuitively change the need for such formalized languages. It may not be a total waste of time, but I fail to see how this slight exaggeration justifies how your comment is currently stamped into the ground, likely by the current livid rage currently directed at LLMs in the software developer community. I get it, people don't like to have their lives disrupted, and I sympathize, just as I sympathized when secretaries, bookstore owners and others were forced to find other ways to make a living. At least in our case we still have (or can find) a job, it has just changed, to some extent even considerably.
ggm 3y ago
My memory of Robert Lacy's biography of Ford is that he also introduced a simplified english for his migrant workers, as well as teaching them english in factory Schools. I'm pretty sure South African miners used simplified english to get around tribal language barriers. (these are distinct from the emergence of pidgin, they're directed intent to create synthetic languages, for specific outcome)
em3rgent0rdr 3y ago
Something like this would make for a great universal human language. In addition to this simplified grammar, it would also be great to limit vocabulary so only need a small dictionary, hopefully without any homonyms, homophones, or homographs.
- mchaver 3y ago
  The best example of this might actually be Wikipedia. There is a separate set of articles in a simpler and controlled form of English: https://simple.wikipedia.org/wiki/Main_Page https://simple.wikipedia.org/wiki/Main_Page , and details about the rules: https://simple.wikipedia.org/wiki/Basic_English https://simple.wikipedia.org/wiki/Basic_English
  - zokier 3y ago
    Simple English Wikipedia is not limited to Basic English
- stavros 3y ago
  I doubt that English is a good basis for a universal anything, due to the extreme inconsistency of spelling and pronunciation. There are some other disadvantages, as there are in any language, but having to learn how a word is spelled and pronounced entirely independently of each other, with no correlation, is a big drain on time. I imagine something like Esperanto would be much more suitable for the role.
  - arp242 3y ago
    The downside of Esperanto is that very few people speak it and is harder to learn in some ways as there's far less media in the language – you got that 1960s William Shatner film that's entirely in Esperanto, and maybe 2 or 3 more, and ... that's it. Meanwhile you can learn English from watching Star Trek, Time Team, SpongeBob, reading Harry Potter, The Hobbit, shitposting on Reddit, etc. For better or worse, English already is the de-facto universal language, or the closest we have to it. We tried the Esperanto thing back in the 20s, and it didn't really work out as there was very little interest outside of a fairly sizeable but ultimately limited group of enthusiasts. I don't see how Esperanto or any other constructed language will take off in the foreseeable future. Continued evolution of English pronunciation and spelling seems like a more realistic and better option – the language will change anyway, as all languages do, so might as well ensure it changes in a reasonable direction. Time spent advocating Esperanto is probably better spent on advocating more consistent English.
zinekeller 3y ago
While not strictly a formal language, Standard Marine Communication Phrases (SMCP, https://wwwcdn.imo.org/localresources/en/OurWork/Safety/Documents/A.918(22).pdf https://wwwcdn.imo.org/localresources/en/OurWork/Safety/Docu...) is also a controlled form of English used in international maritime contexts to minimize ambiguity.
- Animats 3y ago
  That's close to signal flag language.[1] Which makes sense. It's used to convey the same messages, things ships need to communicate to nearby ships. [1] https://en.wikipedia.org/wiki/International_Code_of_Signals https://en.wikipedia.org/wiki/International_Code_of_Signals
imiric 3y ago
Why waste time say lot word when few word do trick? In all seriousness, these are good restrictions to follow when writing user-facing error messages. Developers tend to complicate the language, leading to more confusion.
- junon 3y ago
  ACE isn't for error messages or prose, it's for computers to understand natural language without needing an LLM (my own description). The idea is that ACE defines a strict syntax for English whereby computers can parse it reliably and human can read and write it reliably - just like SQL, but English.
  - MichaelZuo 3y ago
    I eagerly await EQL.
ithkuil 3y ago
While ACE goes far beyond just style, I found that the guidelines set in https://www.plainenglish.co.uk/how-to-write-in-plain-english.html https://www.plainenglish.co.uk/how-to-write-in-plain-english... (and similar attempts) are quite useful and happen to have a lot in common with ACE
pluijzer 3y ago
Not an expert at all but I am curious if training a LLM purely on ACE input would result in a smaller but still capable model. I guess a small regular trained model can then translate ACE to a natural language.
- gwern 3y ago
  Narrowing the task space results in better performance in the short run / with small cheap models, but worse performance in the long run. The model overspecializes and loses any generalizable capabilities because they are not necessary - like the old joke about how professors are like sea squirts, as soon as they get tenure they digest their brains as unnecessary. If you provide the model with adequate scale, the more diverse and challenging the problems, the better (https://gwern.net/scaling-hypothesis#blessings-of-scale https://gwern.net/scaling-hypothesis#blessings-of-scale). So, if you trained a LLM on ACE only (and you somehow had enough to begin with), then it would fail to learn so many things. For example, it wouldn't learn about prose style, or other languages, or different speakers and identities.
rozab 3y ago
I wonder how this parses the classic joke: A programmers spouse sends them to the shops. "Buy a pint of milk. If they have eggs, buy a dozen." The programmer returns with 12 pints of milk.
sillymath 3y ago
I tried to use the demo of the github page, but it doesn't work. It seems the page is stalemate since 2013.
kant2002 3y ago
I like ACE it's very interesting application of Prolog. While it focused on writing logical definitions, I think there interesting way of looking at how we try to express programming languages syntax to newbies in classrooms. I personally think that it is possible to find way of speaking about code which sounds a lot like English. Maybe that's because my English is not adequate. I have super early experiments here: https://kant2002.github.io/EngLang/ https://kant2002.github.io/EngLang/ but it's still super early
JohnDeHope 3y ago
This is weird. I write a lot of technical documentation, and it often comes out reading like this. It's almost like I learned this from somewhere, but didn't realize it.
- ReactiveJelly 3y ago
  Convergent evolution, perhaps
sublinear 3y ago
Does there exist a workplace where product owners have a background in technical writing and they're solely responsible for business logic instead of sloppy requirements?
- fallingmeat 3y ago
  Most high integrity software development includes roles for systems engineers that do this (avionics, medical, nuclear, rail, automotive). See standards like ARP4754A, INCOSE working group on requirements engineering for leads, or if you are really into it, applications of formal methods like DO-333

100 points by sublinear 3y ago

dang 3y ago

Related: Attempto Controlled English - https://news.ycombinator.com/item?id=20950126 https://news.ycombinator.com/item?id=20950126 - Sept 2019 (60 comments) AceWiki:an open-source wiki, that can understand english(almost) - https://news.ycombinator.com/item?id=1563384 https://news.ycombinator.com/item?id=1563384 - July 2010 (1 comment)

oezi 3y ago

In the age of LLMs this seems like a waste of time to even think about constraining and formalizing languages. Why should human users restrict themselves in their expression if LLM systems can easily point out any remaining ambiguities?

ggm 3y ago
What they seem to be doing at the moment is riffing on the ambiguities. If they point them out, then the outcome will be .. less ambiguous language which will tend to be like ACE. Gimme the ball is idiomatic. How do you know it means Give me, and not Furble the ball? We're taught Give me, we say Gimme. Both have a role.
- HPsquared 3y ago
  It could also mean "please give me the management role overseeing the event known as the ball".
tgv 3y ago
I'm interested to see an LLM do that. Technically, nearly every sentence is ambiguous, but that clearly doesn't bother us. An automated system that can point out sentences humans are likely to interpret in multiple ways would be interesting. However, no such thing seems to exist.
- fallingmeat 3y ago
  I recently made an LLM-based tool to identify inconsistencies with the INCOSE guide to writing requirements (identifying vague language, for example). Also have had some success translating those outputs to more structured language like LTL. I think it is doable
- oezi 3y ago
  I am curious why you think this. A basic prompt of pointing out ambiguities should get you quite far. I am pretty sure LLMs would be great to turn regular English into ACE including resolving detected ambiguities.
- euroderf 3y ago
  When working as a tech writer, I was mildly obsessed with this. Nothing drives users away from your documentation like a lack of clarity that signals that (1) you're on your own now, pal, and (2) the writer had zero clue of the potential for chaos in interpreting what he is writing.
stareatgoats 3y ago
You make a valid point, LLMs intuitively change the need for such formalized languages. It may not be a total waste of time, but I fail to see how this slight exaggeration justifies how your comment is currently stamped into the ground, likely by the current livid rage currently directed at LLMs in the software developer community. I get it, people don't like to have their lives disrupted, and I sympathize, just as I sympathized when secretaries, bookstore owners and others were forced to find other ways to make a living. At least in our case we still have (or can find) a job, it has just changed, to some extent even considerably.

ggm 3y ago

My memory of Robert Lacy's biography of Ford is that he also introduced a simplified english for his migrant workers, as well as teaching them english in factory Schools. I'm pretty sure South African miners used simplified english to get around tribal language barriers. (these are distinct from the emergence of pidgin, they're directed intent to create synthetic languages, for specific outcome)

em3rgent0rdr 3y ago

Something like this would make for a great universal human language. In addition to this simplified grammar, it would also be great to limit vocabulary so only need a small dictionary, hopefully without any homonyms, homophones, or homographs.

mchaver 3y ago
The best example of this might actually be Wikipedia. There is a separate set of articles in a simpler and controlled form of English: https://simple.wikipedia.org/wiki/Main_Page https://simple.wikipedia.org/wiki/Main_Page , and details about the rules: https://simple.wikipedia.org/wiki/Basic_English https://simple.wikipedia.org/wiki/Basic_English
- zokier 3y ago
  Simple English Wikipedia is not limited to Basic English
stavros 3y ago
I doubt that English is a good basis for a universal anything, due to the extreme inconsistency of spelling and pronunciation. There are some other disadvantages, as there are in any language, but having to learn how a word is spelled and pronounced entirely independently of each other, with no correlation, is a big drain on time. I imagine something like Esperanto would be much more suitable for the role.
- arp242 3y ago
  The downside of Esperanto is that very few people speak it and is harder to learn in some ways as there's far less media in the language – you got that 1960s William Shatner film that's entirely in Esperanto, and maybe 2 or 3 more, and ... that's it. Meanwhile you can learn English from watching Star Trek, Time Team, SpongeBob, reading Harry Potter, The Hobbit, shitposting on Reddit, etc. For better or worse, English already is the de-facto universal language, or the closest we have to it. We tried the Esperanto thing back in the 20s, and it didn't really work out as there was very little interest outside of a fairly sizeable but ultimately limited group of enthusiasts. I don't see how Esperanto or any other constructed language will take off in the foreseeable future. Continued evolution of English pronunciation and spelling seems like a more realistic and better option – the language will change anyway, as all languages do, so might as well ensure it changes in a reasonable direction. Time spent advocating Esperanto is probably better spent on advocating more consistent English.

zinekeller 3y ago

While not strictly a formal language, Standard Marine Communication Phrases (SMCP, https://wwwcdn.imo.org/localresources/en/OurWork/Safety/Documents/A.918(22).pdf https://wwwcdn.imo.org/localresources/en/OurWork/Safety/Docu...) is also a controlled form of English used in international maritime contexts to minimize ambiguity.

Animats 3y ago
That's close to signal flag language.[1] Which makes sense. It's used to convey the same messages, things ships need to communicate to nearby ships. [1] https://en.wikipedia.org/wiki/International_Code_of_Signals https://en.wikipedia.org/wiki/International_Code_of_Signals

imiric 3y ago

Why waste time say lot word when few word do trick? In all seriousness, these are good restrictions to follow when writing user-facing error messages. Developers tend to complicate the language, leading to more confusion.

junon 3y ago
ACE isn't for error messages or prose, it's for computers to understand natural language without needing an LLM (my own description). The idea is that ACE defines a strict syntax for English whereby computers can parse it reliably and human can read and write it reliably - just like SQL, but English.
- MichaelZuo 3y ago
  I eagerly await EQL.

ithkuil 3y ago

While ACE goes far beyond just style, I found that the guidelines set in https://www.plainenglish.co.uk/how-to-write-in-plain-english.html https://www.plainenglish.co.uk/how-to-write-in-plain-english... (and similar attempts) are quite useful and happen to have a lot in common with ACE

pluijzer 3y ago

Not an expert at all but I am curious if training a LLM purely on ACE input would result in a smaller but still capable model. I guess a small regular trained model can then translate ACE to a natural language.

gwern 3y ago
Narrowing the task space results in better performance in the short run / with small cheap models, but worse performance in the long run. The model overspecializes and loses any generalizable capabilities because they are not necessary - like the old joke about how professors are like sea squirts, as soon as they get tenure they digest their brains as unnecessary. If you provide the model with adequate scale, the more diverse and challenging the problems, the better (https://gwern.net/scaling-hypothesis#blessings-of-scale https://gwern.net/scaling-hypothesis#blessings-of-scale). So, if you trained a LLM on ACE only (and you somehow had enough to begin with), then it would fail to learn so many things. For example, it wouldn't learn about prose style, or other languages, or different speakers and identities.

rozab 3y ago

I wonder how this parses the classic joke: A programmers spouse sends them to the shops. "Buy a pint of milk. If they have eggs, buy a dozen." The programmer returns with 12 pints of milk.

sillymath 3y ago

I tried to use the demo of the github page, but it doesn't work. It seems the page is stalemate since 2013.

kant2002 3y ago

I like ACE it's very interesting application of Prolog. While it focused on writing logical definitions, I think there interesting way of looking at how we try to express programming languages syntax to newbies in classrooms. I personally think that it is possible to find way of speaking about code which sounds a lot like English. Maybe that's because my English is not adequate. I have super early experiments here: https://kant2002.github.io/EngLang/ https://kant2002.github.io/EngLang/ but it's still super early

JohnDeHope 3y ago

This is weird. I write a lot of technical documentation, and it often comes out reading like this. It's almost like I learned this from somewhere, but didn't realize it.

ReactiveJelly 3y ago
Convergent evolution, perhaps

sublinear 3y ago

Does there exist a workplace where product owners have a background in technical writing and they're solely responsible for business logic instead of sloppy requirements?

fallingmeat 3y ago
Most high integrity software development includes roles for systems engineers that do this (avionics, medical, nuclear, rail, automotive). See standards like ARP4754A, INCOSE working group on requirements engineering for leads, or if you are really into it, applications of formal methods like DO-333