- From: Chris Spencer <chrisspen@gmail.com>
- Date: Wed, 18 Mar 2009 13:29:34 -0400
- To: semantic-web@w3.org
Hi, I'm at the brainstorming stage of a semantic application, and I'd like to get some feedback. I'd like to create a web application that would provide a structured debating format that would incorporate semantic rules and logic entered by users. The basic idea is that a user would enter an assertion, evidence, and definitions, and other users would vote on the accuracy of these elements, or extend them with their own arguments. The system would also make an attempt to validate logical arguments, and show where a user's line of reasoning is incomplete, flawed, or contradictory, and would need correction or clarification. To use a controversial political example, a user might enter the assertion "The war in Iraq is a mistake." and then enter the supporting argument "because, it costs billions with no tangible benefit to the US". The user might optionally enter definitions and examples for "tangible benefit" for clarification. Another user might disagree and enter a counter argument, claiming "The Iraqi government shares security intelligence" and that "Security intelligence is a tangible benefit". An arbitrary number of "for" and "against" arguments could be added, and the results accumulated in the parent assertion. The goal isn't to simply find the consensus opinion on an issue. A basic poll can do that. Instead, I'm interesting in finding the clearest most unambiguous and quantifiable line of reasoning supporting an opinion, especially when a subject has conflicting opinions. I used the above example since it has passionate arguments on either side. In these cases, I'm curious to see which position can be quantifiably reduced to the clearest line of reasoning, instead of the knee-jerk oversimplified diatribes we typically see in the main stream media. Another goal would be to filter out misinformation and mischaracterization, using a combination of user votes and automated logic validation. Of course, I realize this is no simple task. I have a little background in machine learning, so I can imagine how impossibly complicated implementing this system could become. I also realize that not everyone's opinion is based on logic, and that a certain amount of "unexplainable bias" is to be expected. However, I'm not too familiar with how much this subject overlaps with the semantic web topic, so my basic questions are: 1. What's the prior art and existing technology? Yes, I've Googled, but since I'm unsure what exact topic this falls under, I don't think I'm getting a comprehensive picture. As far as I can tell, there's nothing that exactly attempts what I'm suggesting. The closet I've found are the various automated logic parsers (e.g. for OWL and RDF), which don't seem to support context, probability, or any kind of accumulation of discrete examples into higher-level statements, which I'd expect this project would require (but please correct me if I'm wrong). 2. Do these goals seem obtainable? Does my idea seem reasonable, insane, far-fetched, etc? Clearly, I've glossed over the technical details, partially for brevity, and partially because I'm not completely sure how I'd implement it. I'd expect the initial system to be limited in the terminology it would support. I doubt you'd be able to enter arguments in natural language. Instead, to help simplify the initial implementation, the user would have to enter a simplified grammar tree. To use the above example, the assertion might be entered in indented form like: The war in Iraq is a mistake because it costs billions with no tangible benefit to the US If such a system were created, I'd imagine it being Wikipedia-like, in that there would be no clear profit motive behind it, so it would have to be open-source and community driven. I appreciate any thoughts. Regards, Chris P.S. Sorry if the above example in any way offends. I'm usually apolitically and generally go out of my way not to advertise my personal political opinions.
Received on Thursday, 19 March 2009 01:34:33 UTC