Join can read and write

Join can feed its output right onto the back end of an input graph

memory_base = plaintext_base join memory_base 

I append the output graph right onto the input graph, both in memory, very high speed. Memory bases in the same database attachment, with a read and write cursor.

 I can load memory base with plain words from plain text. Then those words, begin to organizer into descending groups with specialized meaning. All in the same base. It is a process of bot reading, taking the stream of word from a book, then selectively breaking these words into sub group until the bot has a general to specify graph tree, starting as the great mass of ordered words and ending into specialize, but unordered definitions of the text. A form of word, of text compression by word definition, also known as reading. Every thing I see in data science I see as a queueing problem, reading also.

I bring it up because the simple system test is to strem plain text into memory by joining it with  memory. I can feed stacks of text into the thing and cross it with the stacks of text already retained.

Example:

I have word sets associated with  agriculture, and another collection of words lists associated weih nutrition.  Very quickly join will process the ordered list of text against the various groups, and determine the subject may be food.  e solve one big problem, "Our these thing part of those things and those other things and those other things,. Arem they connected in the stream of ordered words. 

Recursively cross referencing against smaller groups of words, refining the understanding. The graph will rise in rank while group sizes at the nodes become thinner.  The graph becomes thin enough to allow the bot to create meaningful sentences by descending the complete graph.

Th b ot then can say, weityh certainty:
based on prior knowledge of words groups, the solving sentences is likely one of these N sentences, and the bot traverses the graph ten times, following most likely routes.  Join will beat jeopardy/. 

Join does that better and faster than Watson.