AI 3

by Javantea
Oct 2, 2012

All text comes from: Brown Corpus, Wikipedia Database download from 2012-10-01, and Wordnet.

This website was created from scratch by Javantea of AltSci Concepts. It uses simple algorithms to make use of huge databases of free knowledge and parse them into short lists of commands. These commands will control an artificial intelligence. By using the Brown Corpus, Wikipedia, and Princeton's Wordnet (not currently in the database, but used in the natural language interface and the underlying AI), I intend to create an AI that can easily answer the following questions:

[1] When was the Civil War is an incredibly easy question to answer. It is answered by the 15th sentence on the linked page. How do we pick this instead of the others? An algorithm will pick up all date sentences as it trawls through. If the AI is confused by other sentences that contradict, it should either use the most common one (safe bet that popularity is correct) or it can ask "Which Civil War?"

[2] Of these questions, the last is subjective. An AI can answer a subjective question, in fact it has a 50% chance of getting it right by saying yes to everything. We always want to be better than 50% correct though. Given the data found in the link and the AI's specific algorithm to parse the data, an answer can be calculated and the answer will be correct if the algorithm is correct and the data is correct. Different algorithms for different AI would provide different answers, which mimics intelligence quite well. What would be impressive is if it could use the sentences in the link to form an intelligent answer. I believe that a fairly straightforward AI could do this.

The end result of this website will be a filtered version of Wikipedia designed to train AI and humans to use natural language. This will advance the field of artificial intelligence as well as linguistics as we prove that human languages can be parsed like computer languages, only more slowly.

Alpha warningAlpha Warning: This website is in alpha testing, which means that it is very experimental and is likely to break. It does not yet have protection against overzealous users, so be gentle. If you want a full dump of the database, it is far easier for me to just give it to you. All of the data (see the above) is open source and freely available. If you notice the site going slow, it may be because the database is taking too many hits or it could be that it just needs a second to reindex or recache. Don't abuse it during these times.