PDF version of this document
The entry barrier for a project such as this is extremely high. In order to just test an algorithm, vast amounts of data need to be analysed. The more data is available, the better the results should be. Iuron aims at funding from University incubation or possible funding from investors and/or relatives. Here are a few possible funding frameworks, listed in order of precedence or likelihood of success.
It is worth re-iterating and stressing the fact that Iuron aspires to be somewhat of a knowledge base, hence it must exploit an abundance of information and sophisticated machine learning approaches rather than simple word matching, counting, and storage. Consequently, it needs high funds to become viable or even be possible to argue about as successful (and verging a point of empirical positive evidence). Due to scale that is required to make our statistical sample for learning sufficient, ad-hoc methods must be devised, particularly at the start.
We may take a genetic algorithms approach, whereby weak facts are discouraged and repetitions are perceived as encouraging. Larger scale will lead to more accuracy, saturation and reliability. This method may also be rather immune to spam unlike some traditional searches, but PageRank-like mechanisms still need to be put in place.
It is important to consider conflicting interest and deceiving knowledge sources that use repeatable false content (``spam'' where its meaning become ``mass lies''). For example, a pharmaceutical company will have financial incentive in spreading a false word, according to which their drug is the best solution to an illness of some kind.
Trust is extremely important, much like attempts at TrustRank and human moderation at DMOZ (a non-profit Web directory) have shown. DMOZ gets a positive mention as opposed to Yahoo's corporate-inclined directory where money warrants listing. Another problem with PageRank is that ranks can be purchased in the form of link. So, power and influence can be bought rather than rightfully earned. Due to the scale of the Internet, fraud cannot be controlled by a human. The system must be self-sustaining. Algorithm secrecy and obscurity is often the vendor's solution to the issue. Comment spam, ``Googlebombing'' and the like are some of the detrimental by-products of a deficient algorithm.
These challenges or barriers cannot be trivially solved by this paradigm which is knowledge engine. However, its extra complexity should open more doors to optimisation, refinement, and improvements.