- spec by clause
	- dependent on other clauses
test cawmp with different algs

- ontol to specs

- qa ask for data structures based on computational verb
	- store in dictionary - many cvs to one spec (like t2b)
- data (spec) generation algs write data based on clues from nouns, verb
- data is passed from one alg part to another and more specs are generated

* connect (all simple, multiple) dict specs together
x first spec line is broken into hierarchies of specs

- first thought from dance, second thought from dance, data are defined in comments, new spec - sentences are analysed to find data structures, spec
- use "Please fill in." for unknown data
- if new language, identify separate algorithm in one of the steps

- use label as well as spec
- separate spec tester and dictionary algs x find alg parts from specs (with queries for separate parts to test), then save for modification
	- functionalism for lists of lists, etc. - do later
- words like *wildcard, list (can have synonyms and) trigger certain structures
	- x use texttoalg with previous line *
	
for b, identify new fail case and for b solution, simplify data
- if the above sentence was an alg sent, then would find b alg and apply fail case (an alg operation that simplifies data)
- b is a class of algs with name [b,*]
	 - 10 Bs which apply to content (e.g. if B to word "content", then insert a mistake)
	 - e.g. B not x
	 
	- * B is also in alg dict
	- templates x words for two uses, list recursion
	
	- br out random algs ** (if ** was an alg sent - [random_algs, br], where the ontol/grammar matches POS after asking for verb, noun, 
	chooses furthest apart n, v pair x separate lists)


10 random data items per list of same type
avoid data repeating in list
b replaces with other data of same type
- to generate detail sentences, replace with own brdict2 word's brdict1 word or n with box or obj, v with right and a with plus

- x use (start of gla with x following takes care of sentence keywords) algwriter, taking care of b, b sols and details