Thoughts, Symbols, and Neurons

Mindmeld is made up primarily of thoughts, symbols, and neurons.
Memory Diagram

A thought is a specific answer to a specific question, and includes a summary (the question being answered) and a detail (the answer itself, in text or HTML format). To simplify the example, we'll only use the summary and ignore the detail.
A symbol is an alphanumeric term: a word, a number or some combination of the two. Mindmeld stores one copy of each unique symbol that ever occurs in a thought or is asked during a search.
Examples of symbols: DOG, ERROR1274, 256
A neuron links a single thought with a single symbol. Each neuron has a strength that describe how relevant the symbol is to that particular thought. Each thought and each symbol can have many neurons.

Searching for Answers

What happens when you submit "My dog is cold" to Mindmeld?

First, Mindmeld breaks the question into search symbols, or stimuli.

Next, Mindmeld identifies each symbol in its database that matches one of the stimuli, and lists the neurons linking those symbols to thoughts. For every thought returned, the strength of its neurons that took part in the search is summed, resulting in the final weight of that thought. You are presented with a list of the top thoughts that matched your search, sorted by weight. Query Diagram

Finally, you view the answer and provide feedback. Here's what happens when you click the "YES" (this answer was helpful) button:

Each neuron that in the viewed answer that took part in the search gets positively reinforced, increasing its strength.

Strengthened neurons for this thought:	MY	IS	COLD

Each neuron in that answer that didn't take part in the search gets negatively reinforced slightly, decreasing its strength.

Weakened neurons for this thought:	GREYHOUND

Each neuron that took part in the search but isn't attached to the selected answer gets negatively reinforced slightly, decreasing its strength.

Weakened neurons for other thoughts:	MY	IS

Each stimulus that doesn't exist as a symbol gets added as a new symbol and gets linked to the selected answer.

New Symbols added and linked to this thought:	DOG

Mindmeld has learned that DOG is relevant to this thought.

But what happens when you select the "NO" (this answer wasn't helpful) button?

Each neuron that in the viewed answer that took part in the search gets negatively reinforced slightly, decreasing its strength.

Weakened neurons for this thought:	MY	IS	COLD

What's it all Mean?

First of all, it means Mindmeld can learn new words and relate them to answers dynamically. Further, it can determine the importance of a given word with respect to a specific answer. It adapts. It is Borg.

Over time, Mindmeld can identify common synonyms and mispellings automatically. This follows from the previous point; if you misspell a word, but eventually get to the right answer, Mindmeld will add the misspelled word as a symbol.

Frequently-used answers bubble to the top, infrequently-used answers require more specific searches to find. Just like your own memory.

Noisewords and All That

What about "MY" and "HOW DO I" and all those other generally meaningless words that are getting strengthened during the searches?

Mindmeld uses basic statistical analysis to identify words that occur so frequently as to be meaningless in a search. Words such as "A," "MY," etc., exist as symbols, but are generally ignored during a search. Note: The noise filter will not begin marking noisewords until there's enough data (from answers and searches) in your knowledgebase to be statistically significant.