An idea like no other... I need help!
-
I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. This program could use the words originally typed in as well as new ones. For example, if one typed in: "Sea Spiders are only found at the ocean floor, where there is little to no sunlight." It could rearrange the sentence into something like this: "Sea Spiders are only located where there is barely sunlight; the ocean floor." I understand it is difficult for computers to understand written language, which is why I'm unsure if this is even possible. I know that the particular bot/program I'm thinking of must do the following: -Identify sentence type -Identify Subject and Verb -Find synonyms for verb, and use them appropriately -rearrange sentence -grammar check sentence -correct grammar -Return finished product. From what I've gathered from various forums and tech whizzes, my idea is certainly plausible, however, it is very difficult. After looking at several bots that can comprehend what you say to it and then return an answer that makes sense, I feel like what I'm talking about should not be that as hard as everyone makes it out to be. As far as program design, I guess I would just work on one small thing at a time. Sentence rewriter? Too vague. Change some words? Still too vague. Change all adjectives with their synonyms? That's a small enough task, but vague in programming terms. Use a function to search the sentence for adjectives against a huge database file of adjectives, then create a loop to take each adjective and replace it with another synonymous adjective in a database file? Now that is both specific enough in human and programming terms. I suppose I should just break down each task in human terms first, then take each task and write it out in programming terms. I need help and input though, thanks a lot!!
-
I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. This program could use the words originally typed in as well as new ones. For example, if one typed in: "Sea Spiders are only found at the ocean floor, where there is little to no sunlight." It could rearrange the sentence into something like this: "Sea Spiders are only located where there is barely sunlight; the ocean floor." I understand it is difficult for computers to understand written language, which is why I'm unsure if this is even possible. I know that the particular bot/program I'm thinking of must do the following: -Identify sentence type -Identify Subject and Verb -Find synonyms for verb, and use them appropriately -rearrange sentence -grammar check sentence -correct grammar -Return finished product. From what I've gathered from various forums and tech whizzes, my idea is certainly plausible, however, it is very difficult. After looking at several bots that can comprehend what you say to it and then return an answer that makes sense, I feel like what I'm talking about should not be that as hard as everyone makes it out to be. As far as program design, I guess I would just work on one small thing at a time. Sentence rewriter? Too vague. Change some words? Still too vague. Change all adjectives with their synonyms? That's a small enough task, but vague in programming terms. Use a function to search the sentence for adjectives against a huge database file of adjectives, then create a loop to take each adjective and replace it with another synonymous adjective in a database file? Now that is both specific enough in human and programming terms. I suppose I should just break down each task in human terms first, then take each task and write it out in programming terms. I need help and input though, thanks a lot!!
So, put another way, you want to make a tool to help someone plagiarize content. :)
-
So, put another way, you want to make a tool to help someone plagiarize content. :)
-
I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. This program could use the words originally typed in as well as new ones. For example, if one typed in: "Sea Spiders are only found at the ocean floor, where there is little to no sunlight." It could rearrange the sentence into something like this: "Sea Spiders are only located where there is barely sunlight; the ocean floor." I understand it is difficult for computers to understand written language, which is why I'm unsure if this is even possible. I know that the particular bot/program I'm thinking of must do the following: -Identify sentence type -Identify Subject and Verb -Find synonyms for verb, and use them appropriately -rearrange sentence -grammar check sentence -correct grammar -Return finished product. From what I've gathered from various forums and tech whizzes, my idea is certainly plausible, however, it is very difficult. After looking at several bots that can comprehend what you say to it and then return an answer that makes sense, I feel like what I'm talking about should not be that as hard as everyone makes it out to be. As far as program design, I guess I would just work on one small thing at a time. Sentence rewriter? Too vague. Change some words? Still too vague. Change all adjectives with their synonyms? That's a small enough task, but vague in programming terms. Use a function to search the sentence for adjectives against a huge database file of adjectives, then create a loop to take each adjective and replace it with another synonymous adjective in a database file? Now that is both specific enough in human and programming terms. I suppose I should just break down each task in human terms first, then take each task and write it out in programming terms. I need help and input though, thanks a lot!!
zf222 wrote:
I feel like what I'm talking about should not be that as hard as everyone makes it out to be.
Wanna bet?? Google "Turing Test" and find out just how hard that really is. Just throwing a thesaurus or a synonym dictionary at a sentence doesn't really do the job. There are subtle differences in the meanings of certain words that would just make them sound really funky in the wrong context.
A guide to posting questions on CodeProject[^]
Dave Kreskowiak Microsoft MVP Visual Developer - Visual Basic
2006, 2007, 2008modified on Friday, November 14, 2008 10:09 AM
-
So, put another way, you want to make a tool to help someone plagiarize content. :)
-
Robert Surtees wrote:
So, put another way, you want to make a tool to help someone plagiarize content. Smile
If that's what you were going to use it for.. then yes, I suppose it would help you get away with a few things.
It's a huge undertaking. This[^] might be of help. I would think the difficulty would be similar to language translation (which is done very poorly) x1000. Here's the result of your first two sentences put through Babelfish twice. English -> Spanish -> English: I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. becomes I want to create a tool of the writing, something that could take an oration and rewrite it in a diverse oration that means just as first. It thinks about her like treasure of the oration.
modified on Wednesday, November 12, 2008 10:48 PM
-
zf222 wrote:
I feel like what I'm talking about should not be that as hard as everyone makes it out to be.
Wanna bet?? Google "Turing Test" and find out just how hard that really is. Just throwing a thesaurus or a synonym dictionary at a sentence doesn't really do the job. There are subtle differences in the meanings of certain words that would just make them sound really funky in the wrong context.
A guide to posting questions on CodeProject[^]
Dave Kreskowiak Microsoft MVP Visual Developer - Visual Basic
2006, 2007, 2008modified on Friday, November 14, 2008 10:09 AM
Dave Kreskowiak wrote:
There are subtle differences in the meanings of certain words that would just make them sound really funky in the wrong context.
Most humans aren't capable of understanding that fact. How will we ever teach a computer to do it? I have a ;arge and versatile vocabulary, and I use it freely. It frustrates me immeasurably that 99% of the people I've known are completely incapable of understanding the subtle nuances of what I most precisely say to them. Conversely, I really appreciate the few who do. :-D
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
-
I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. This program could use the words originally typed in as well as new ones. For example, if one typed in: "Sea Spiders are only found at the ocean floor, where there is little to no sunlight." It could rearrange the sentence into something like this: "Sea Spiders are only located where there is barely sunlight; the ocean floor." I understand it is difficult for computers to understand written language, which is why I'm unsure if this is even possible. I know that the particular bot/program I'm thinking of must do the following: -Identify sentence type -Identify Subject and Verb -Find synonyms for verb, and use them appropriately -rearrange sentence -grammar check sentence -correct grammar -Return finished product. From what I've gathered from various forums and tech whizzes, my idea is certainly plausible, however, it is very difficult. After looking at several bots that can comprehend what you say to it and then return an answer that makes sense, I feel like what I'm talking about should not be that as hard as everyone makes it out to be. As far as program design, I guess I would just work on one small thing at a time. Sentence rewriter? Too vague. Change some words? Still too vague. Change all adjectives with their synonyms? That's a small enough task, but vague in programming terms. Use a function to search the sentence for adjectives against a huge database file of adjectives, then create a loop to take each adjective and replace it with another synonymous adjective in a database file? Now that is both specific enough in human and programming terms. I suppose I should just break down each task in human terms first, then take each task and write it out in programming terms. I need help and input though, thanks a lot!!
Have you ever looked into the old language, Prolog? A lot of work was done in that language on Natural Language Processing. I played with it, but never mastered it, because its conceptual base is so totally foreign to my way of approaching problems. But I've always believed that if a child was introduced to it before sequential thinking was canalized in its brain, that language would be far more natural to the way the brain really works than anything else we've developed. It's based on predicate logic, and is ideally suited to parallel processing techniques, which would probably be valuable in a tool such as you describe. The Inference Engine is intended to follow multiple paths through a set of rules, backtracing when a match cannot be found on a given path, until all known rules are exhausted. It can be extended to learn new rules, and to apply fuzzy logic to assign probabilities to certain outcomes. I suspect it would be a perfect match for your application with a lot less work than trying to implement it in conventional languages. Admittedly, I haven't really played with it since Turbo Prolog 2.0 for DOS, but it was far advanced then, and I understand that it has still got a small following trying to develop it further. I encourage you to give it a look. I'm a big fan of what the creators were trying to do with it, but I'm much too stupid to go much beyond that. Better luck to you... :-D
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
-
I want to create a writing tool, something that could take a sentence and rewrite it into a different sentence that means the same as the first. Think of it as a sentence thesaurus. This program could use the words originally typed in as well as new ones. For example, if one typed in: "Sea Spiders are only found at the ocean floor, where there is little to no sunlight." It could rearrange the sentence into something like this: "Sea Spiders are only located where there is barely sunlight; the ocean floor." I understand it is difficult for computers to understand written language, which is why I'm unsure if this is even possible. I know that the particular bot/program I'm thinking of must do the following: -Identify sentence type -Identify Subject and Verb -Find synonyms for verb, and use them appropriately -rearrange sentence -grammar check sentence -correct grammar -Return finished product. From what I've gathered from various forums and tech whizzes, my idea is certainly plausible, however, it is very difficult. After looking at several bots that can comprehend what you say to it and then return an answer that makes sense, I feel like what I'm talking about should not be that as hard as everyone makes it out to be. As far as program design, I guess I would just work on one small thing at a time. Sentence rewriter? Too vague. Change some words? Still too vague. Change all adjectives with their synonyms? That's a small enough task, but vague in programming terms. Use a function to search the sentence for adjectives against a huge database file of adjectives, then create a loop to take each adjective and replace it with another synonymous adjective in a database file? Now that is both specific enough in human and programming terms. I suppose I should just break down each task in human terms first, then take each task and write it out in programming terms. I need help and input though, thanks a lot!!
I'm working on a word prediction application and we have some common problems. I don't think that looking at you application as one of translation is a good approach. Translation is made with losses. English is a low-inflected language and that is a big advantage for you. I'm thinking at your application as a semantic-syntactic prediction problem. Read Text prediction systems: a survey[^] and see the biography. . You can find the pdf version(16 pages) here[^].