9b4a197 · opencog-agents

Changed files (1) +40 -18

scm/pair-agent/count-agent.scm +40 -18

…	…	@@ -9,11 +9,14 @@
9	9	; This source blocks if there's nothing to be read; else it
10	10	; returns the next PhraseNode. The SRFI's call this a generator.
11	11	; For now, see `sensory.scm` module and example.
	12	; It uses a FileReadNode to loop voer a text file.
12	13	; 2) (Optional) Some way to split one source into multiple sources.
13	14	; Maybe this:
14	15	; https://wiki.opencog.org/w/PromiseLink#Multiplex_Example
15	16	; but what happens if the readers don't both read ???
16	17	; 3) A Filter that takes above and increments pair counts.
	18	; Done. Its non-atomic, but so what. Perhaps we can live with
	19	; that, given that sampliing is statistical, anyway.
17	20	; 4) Execution control. There are two choices:
18	21	; Pull: infinite loop polls on promise. Source blocks if
19	22	; no input.
…	…	@@ -22,22 +25,32 @@
22	25	; Right now, the general infrastructure supports Pull naturally.
23	26	; There aren't any push demos.
24	27	; Attention control is easier with Pull.
	28	;
	29	; The below demos this design.
	30	;
	31	; Open issues:
	32	; *) Need to call cog-execute! in a loop to loop over whole file.
	33	; Would be nicer to not have to do this.
	34	; *) End-of-file results in a throw from parser. This is a side
	35	; effect. We need some proper end-of-file handling.
25	36
26
27		(define text-blob "this is just a bunch of words you know")
28
29
	37	; --------------------------------------------------------------
30	38	; Return a text parser that counts edges on a stream. The
31	39	; `txt-stream` must be an Atom that can serve as a source of text.
32	40	; Typically, `txt-stream` will be
33	41	; (ValueOf (Concept "some atom") (Predicate "some key"))
34	42	; and the Value there will be a LinkStream from some file or
35	43	; other text source.
	44	;
	45	; These sets up a processing pipeline in Atomese, and returns that
	46	; pipeline. The actual parsing all happens in C++ code, not in scheme
	47	; code. The scheme here is just to glue the pipeline together.
36	48	(define (make-parser txt-stream)
37		; From inside to out:
38		; LGParseBonds tokenizes a sentence, and then parses it.
39		; The PureExecLink makes sure that the parsing is done in a sub-AtomSpace
40		; so that the main AtomSpace is not garbaged up.
	49	;
	50	; Pipeline steps, from inside to out:
	51	; * LGParseBonds tokenizes a sentence, and then parses it.
	52	; * The PureExecLink makes sure that the parsing is done in a
	53	; sub-AtomSpace so that the main AtomSpace is not garbaged up.
41	54	;
42	55	; The result of parsing is a list of pairs. First item in a pair is
43	56	; the list of words in the sentence; the second is a list of the edges.
…	…	@@ -46,8 +59,13 @@
46	59	; (LinkValue (Word "this") (Word "is") (Word "a") (Word "test"))
47	60	; (LinkValue (Edge ...) (Edge ...) ...))
48	61	;
49		; The Filter matches this, so that (Variable "$x") is equated with the
50		; list of Edges.
	62	; The outer Filter matches this, so that (Glob "$edge-list") is
	63	; set to the LinkValue of Edges.
	64	;
	65	; The inner Filter loops over the list of edges, and invokes a small
	66	; pipe to increment the count on each edge.
	67	;
	68	; The counter is a non-atomic pipe of (SetValue (Plus 1 (GetValue)))
51	69	;
52	70	(define NUML (Number 4))
53	71	(define DICT (LgDict "any"))
…	…	@@ -86,9 +104,9 @@
86	104	(Type 'LinkValue) ; the word-list
87	105	(LinkSignature ; the edge-list
88	106	(Type 'LinkValue) ; edge-list wrapper
89		(Glob "$x"))) ; all of the edges
90		; Rewrite
91		(count-edges (Glob "$x"))
	107	(Glob "$edge-list"))) ; all of the edges
	108	; Pipeline to apply to the resulting match.
	109	(count-edges (Glob "$edge-list"))
92	110	)
93	111	(parseli phrali)))
94	112
…	…	@@ -96,8 +114,10 @@
96	114	(filty txt-stream)
97	115	)
98	116
99		; Parse a string.
	117	; Demo wrapper: Parse a string.
100	118	(define (obs-txt PLAIN-TEXT)
	119
	120	; We don't need to create this over and over; once is enough.
101	121	(define txt-stream (ValueOf (Concept "foo") (Predicate "some place")))
102	122	(define parser (make-parser txt-stream))
103	123
…	…	@@ -110,8 +130,10 @@
110	130	(cog-extract-recursive! phrali)
111	131	)
112	132
113		; Parse contents of a file.
	133	; Demo wrapper: Parse contents of a file.
114	134	(define (obs-file FILE-NAME)
	135
	136	; We don't need to create this over and over; once is enough.
115	137	(define txt-stream (ValueOf (Concept "foo") (Predicate "some place")))
116	138	(define parser (make-parser txt-stream))
117	139
…	…	@@ -148,7 +170,7 @@
148	170	; risks deadlock.
149	171	;
150	172	; cog-update-value calls
151		; asp->increment_count(h, key, fvp->value()));
	173	; Ideally, call asp->increment_count(h, key, fvp->value()));
152	174
153	175	; Below works but is a non-atomic increment.
154	176	(define ed (Edge (Bond "ANY") (List (Word "words") (Word "know"))))
…	…	@@ -158,7 +180,7 @@
158	180	(Plus (Number 0 0 1)
159	181	(FloatValueOf ed (Predicate "count") (Number 0 0 0))))))
160	182
161		; Lets work on file-access now.
	183	; Here's a demo of what file-access looks like.
162	184	(cog-execute!
163	185	(SetValue (Concept "foo") (Predicate "some place")
164	186	(FileRead "file:///tmp/demo.txt")))
…	…	@@ -170,7 +192,7 @@
170	192	; (cog-execute! (LgParseBonds txt-stream-gen DICT NUML))
171	193
172	194	; (load "count-agent.scm")
173		; (obs-txt text-blob)
	195	; (obs-txt "Some kind of sentence to process, hell yeah!")
174	196	; (obs-file "/tmp/demo.txt")
175	197	; (cog-report-counts)
176	198	; (cog-get-atoms 'WordNode)

opencog-agents commits / 9b4a197

Changed files (1) +40 -18