tursiops, what kind of poems would this generate? Modern free verse, or rhymed and metered poems, sonnets, or something else? And what would the tone of them be? (Or would that depend on the dictionary of words you fed into it?)
In fact this is very open, and the idea very general. I was thinking of maybe choose one sort of poem at the beginning, and then add more options later.
If you want to see what kind of poems I write myself, fo to tursiops33.deviantart.com/
Ok, I’ve pledged some more for my project because I really trust in it! I also bough the domain name thepoemgenerator.com and redirected to this page.
Just waiting for a quote from a nice developper ^^
This is impossible for the following reason: “It will be a real webapp that understand grammar”
It is impossible to create a computer program that understands grammar. If someone knew how to do this they would be making hundreds of thousands of dollars per year, if not more.
It is possible to make a poem generator that is slightly passable, but you’re going to end up with cookie cutter templates with different words/adjectives/adverbs/verbs/nouns being used. It can do templates, but there is no way for a program to truly understand grammar.
In fact it will be like the automatical correction in programs, or more like this program (in french) antidote.
No program will ever be as good as a human comprehension but at least we can try. And that’s what I going to do.
Thank you for your comments.
It certainly will not be as good as a human one, but I think that it will be good enough for people who just want to have something quick and unique.
This project idea is fun and I would like to work on it. Programming an AI that generates grammatically correct sentences (even if they’re bogus) is easy - making sure that they rhyme is a little more complicated, but feasible.
Mind you: most of the poems will be quite dadaistic. They will be poems alright, with a proper structure, rhyme, meter, etc. But the computer will have no clue whatsoever what those words mean.
You can make it easy by defining certain contexts - for instance, a love poem would contain words like “kiss”, “love”, “heart”, etc. But still, don’t expect Shakespeare quality.
(I see this as a fun project, but I do have to live of something and I do have to use my AI expertise - that and the dollar is frightfully low compared to the Euro - that’s why I quoted it for $500 instead of the $200 you already pledged.)
You know, let’s set up a website with the generator once it’s finished. I’ll host the service, you pay for the domain and we put up tip jars, adsense and links to shops where you can print your favourite poem on a shirt - and we split the profit. What do you say?
That would take some pressure from both of us (because even if you bring up the $500 - I guess nobody else will pledge on it and it’s a lot of money for one man and a toy project - it still means that I’m working for less than I need).
I can save you guys a bit of trouble, something for your purposes already exists, and it’s GPLed (I believe). Its called the Dada Engine. I’m not sure if i can put link here, but you can definitely find it on google. dev.null.org/dadaengine/
You just write a script, can be simple or incredibly complex. You can just send me the $200 or give it to the guy who wrote it… i’m sure he’d accept a check.
And for the guy at the top that said a computer could never understand grammar, well it doesn’t really have to, it just needs rules, but I’ll bet it makes less mistakes than me or you.
Thanks for the pointer, I know the dada engine and I had a look at it before.
The thing about that is: yes, it goes in the direction of what one needs for that system, but it’s only half the way down the road.
For instance: with the engine it’s not easy to define meters or rhyme, there’s no corrective measure for poems that are supposed to make sense.
So you either have to restrict the scope of the poems heavily (which is not what tursiops wants) or the poems will be indeed very dadaistic. But you can restrict that, which is what I’m doing.
I see that my project has interested a few people, and I’m happy about that.
I wanted this project to be GPLed so that anyone can contribute to it.
If we can upgrade the dada program, then that’s fine for me.
If this program cannot be used and we need to start again, then we’ll do.
I’ll still need to think of it, on what is the best way for this to be the most successful and efficient possible.
Any other comments are greatly appreciated!
Add a comment
Before you add a comment you must be signed up – it takes about 30 seconds. Sign up now.
17 comments (oldest first)
tursiops, what kind of poems would this generate? Modern free verse, or rhymed and metered poems, sonnets, or something else? And what would the tone of them be? (Or would that depend on the dictionary of words you fed into it?)
In fact this is very open, and the idea very general. I was thinking of maybe choose one sort of poem at the beginning, and then add more options later. If you want to see what kind of poems I write myself, fo to tursiops33.deviantart.com/
Ok, I’ve pledged some more for my project because I really trust in it! I also bough the domain name thepoemgenerator.com and redirected to this page. Just waiting for a quote from a nice developper ^^
This is impossible for the following reason: “It will be a real webapp that understand grammar”
It is impossible to create a computer program that understands grammar. If someone knew how to do this they would be making hundreds of thousands of dollars per year, if not more.
It is possible to make a poem generator that is slightly passable, but you’re going to end up with cookie cutter templates with different words/adjectives/adverbs/verbs/nouns being used. It can do templates, but there is no way for a program to truly understand grammar.
In fact it will be like the automatical correction in programs, or more like this program (in french) antidote. No program will ever be as good as a human comprehension but at least we can try. And that’s what I going to do.
@raize-you’re not entirely correct, it is possible for a program to ‘understand’ grammar. however, a webapp doing so is slightly more difficult.
@tursiops-what makes poetry good most of the time is the human element-take that out, and i’m not so sure you’ll like what you get. just a thought.
Thank you for your comments. It certainly will not be as good as a human one, but I think that it will be good enough for people who just want to have something quick and unique.
This project idea is fun and I would like to work on it. Programming an AI that generates grammatically correct sentences (even if they’re bogus) is easy - making sure that they rhyme is a little more complicated, but feasible.
Mind you: most of the poems will be quite dadaistic. They will be poems alright, with a proper structure, rhyme, meter, etc. But the computer will have no clue whatsoever what those words mean.
You can make it easy by defining certain contexts - for instance, a love poem would contain words like “kiss”, “love”, “heart”, etc. But still, don’t expect Shakespeare quality.
(I see this as a fun project, but I do have to live of something and I do have to use my AI expertise - that and the dollar is frightfully low compared to the Euro - that’s why I quoted it for $500 instead of the $200 you already pledged.)
Thank you for your proposal. I’ll try to get the missing money asap, so that you can start to work on it.
You know, let’s set up a website with the generator once it’s finished. I’ll host the service, you pay for the domain and we put up tip jars, adsense and links to shops where you can print your favourite poem on a shirt - and we split the profit. What do you say?
That would take some pressure from both of us (because even if you bring up the $500 - I guess nobody else will pledge on it and it’s a lot of money for one man and a toy project - it still means that I’m working for less than I need).
Thank you for the proposition, but I already have my own plans for this project. This will be released as GPL licence.
Ok, fair enough.
Yes, of course it will be GPLed - but that doesn’t restrict the uses.
I can save you guys a bit of trouble, something for your purposes already exists, and it’s GPLed (I believe). Its called the Dada Engine. I’m not sure if i can put link here, but you can definitely find it on google. dev.null.org/dadaengine/
You just write a script, can be simple or incredibly complex. You can just send me the $200 or give it to the guy who wrote it… i’m sure he’d accept a check.
And for the guy at the top that said a computer could never understand grammar, well it doesn’t really have to, it just needs rules, but I’ll bet it makes less mistakes than me or you.
Hey I’ll write you a Bukowski poem generator with it for $10!
Hi, holden!
Thanks for the pointer, I know the dada engine and I had a look at it before.
The thing about that is: yes, it goes in the direction of what one needs for that system, but it’s only half the way down the road.
For instance: with the engine it’s not easy to define meters or rhyme, there’s no corrective measure for poems that are supposed to make sense.
So you either have to restrict the scope of the poems heavily (which is not what tursiops wants) or the poems will be indeed very dadaistic. But you can restrict that, which is what I’m doing.
I see that my project has interested a few people, and I’m happy about that. I wanted this project to be GPLed so that anyone can contribute to it. If we can upgrade the dada program, then that’s fine for me. If this program cannot be used and we need to start again, then we’ll do. I’ll still need to think of it, on what is the best way for this to be the most successful and efficient possible. Any other comments are greatly appreciated!
Add a comment
Before you add a comment you must be signed up – it takes about 30 seconds. Sign up now.