KWSentence – the original kweb bios text has been parsed into sentences, in order to use those sentences to give context to the crucially important “connections” (links) represented in those sentences. I think this is a crucial “half-way house” form of knowledge, somewhere between structured facts...
More
KWSentence – the original kweb bios text has been parsed into sentences, in order to use those sentences to give context to the crucially important “connections” (links) represented in those sentences. I think this is a crucial “half-way house” form of knowledge, somewhere between structured facts and full text descriptions.
Each sentence has properties linking to its preceding and following sentences, so that a UI could be constructed which enables a variable amount of “contextual text” to be displayed for each connection.
All sentences are parsed for obvious mention of dates, and these dates are stored as properties of the sentence. I have chosen to represent all dates as /measurement_unit/time_interval objects, as most dates are in fact time periods with a beginning and an end. Even the fundamental Freebase /type/datetime acknowledges this by being able to represent different granularities such as a particular day in history, a month in a year, or a whole year. However, many of kweb's time periods will be things like “March to June 1868”, “from the 20th March to the 16th April”, or 1766-89, or even “the next few weeks” (assuming some starting point) or “the summer of that year”. Hence it seems appropriate to use explicit time intervals for everything. This of course generates anachronisms such as “1868” being represented as a redundant) time interval with both start and end dates being 1868, and the same for individual dates. I welcome comments on this.
Sentences also have links to the KWConnections they contain (if any).
Less