Next: FEASIBILITY
Up: Iuron.com Manifestation - Initial
Previous: FUNDING AND SUPPORT
At a rather shallow level, the approach can be outlined as listed
below:
- Strip pages from tags (further help is available in the newsgroups)
- Use headers and tags to highlight important facts
- Identify synonyms (language becomes an issue, but maybe translation
can bridge the gap)
- Collect a list of facts from the page/s in question
For data reliability, we may consider using Wikipedia as a better
facts source where mutual moderation is perpetually forced. The grand
scheme is to crawl pages and not to index and summarise them, but
rather to accumulate knowledge, much like a human reader would do.
It is always worth remembering (caching) sources of information to
refer the reader back to. This would establish confidence and further
breadth for the user's mind. Better priority should be given to pages
with stronger PageRank et al., i.e. pages with more inbound links.
Moreover, it is worth using age of domains, professional affiliation
and so forth as factors; all of these are also worth scoring accumulatively.
Impact should be emphasised as an important aspect in oder
to avoid false facts from ever being absorbed as truthful ones.
As for the user's side, voting mechanism can be used by the engines
or even explicit queries made in natural language and then interpreted
logically (first order predicates). For example, the user can ask
a question or provide some query terms. He/she will consequently get
answers sorted by certainty of response/answer with relevant links/pointer
to the sources; snippets as well can be attached to answers if cache
is available to access.
Next: FEASIBILITY
Up: Iuron.com Manifestation - Initial
Previous: FUNDING AND SUPPORT
Roy Schestowitz
2005-10-12