Talk:Tiki forum archive/Forum 5/8588
Source: 2013 Tiki forum 5, thread #8588. Original authors, UTC times, post IDs, and reply-parent IDs are preserved. This is an archive, not an active article discussion.
Parsing NIhO sections of text
Parsing NIhO sections of text
brablonau · 2009-12-03 11:39 UTC · Tiki post #8588
I'm trying to parse out sections of Lojban text delimited by sequences of NIhO cmavo into their respective paragraphs, sections, chapters, etc.
So, if I have:
ni'o ni'o broda ni'o broda ni'o ni'o broda
I would like to get something like:
[[broda,broda],[broda]]
where the inner brackets represent paragraphs, the outer brackets represent sections, and further containing brackets would designate chapters, parts, volumes, etc.
I'm trying to use the DCG facilities of Prolog to do this. For simplicity, I'm using "p" to represent a paragraph and "n" to represent a cmavo from NIhO.
The CLL states that a text utilizing NIhO should start with a string of NIhOs as long as any other NIhO string in the text. I managed to create grammar rules to parse paragraph structure, AS LONG AS the above condition is met. The following DCG clauses do this well:
parse(0,p) --> [p]. parse(_,[]) --> []. parse([O|N],[H|T]) --> [n], parse(N,H), {H \= []}, parse([O|N],T).
They find the correct parse, and only the correct parse. (i.e., backtracking always terminates and never finds any more solutions.)
Here's an example of the parser in action:
| ?- phrase(parse(Depth,Parse),[n,n,p,n,p,n,n,p]).
Depth = [_,_|0] Parse = [[p,p],[p]] ?
This is the same structure as in the {broda} example above. (Note: the Depth returned is in Peano form: 1 = [_|0], 2 = [_,_|0], etc.)
The problem I'm having is that when the CLL condition is NOT met... that is, when a longer string of NIhOs appears somewhere down the line, the text will fail to parse. For example:
| ?- phrase(parse(Depth,Parse),[n,n,p,n,n,n,p,n,n,p]).
no
That "no" is Prolog's way of saying that the phrase "[n,n,p,n,n,n,p,n,n,p]" doesn't satisfy the grammar defined for "parse". That's because the text starts with NIhO NIhO (two NIhOs) but has NIhO NIhO NIhO (three NIhOs) further along in it.
I've tried two different approaches, now, to infer how many NIhOs are missing at the front of the text. One of these approaches worked correctly, but it required the use of cuts(!) to prevent infinite recursion. While that's fine for parsing, using cuts really is cheating when it comes to writing grammar rules.
Does anyone here know how I could use contetx-free grammar rules to parse the different sections separated by NIhO sequences?
Any ideas (expressed in EBNF, Prolog, YACC, or whatever you speak) would be much appreciated!
ki'e
mi'e brablonau
To unsubscribe from this list, send mail to [email protected]
with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if
Parsing NIhO sections of text
Minimiscience · 2009-12-03 11:39 UTC · Tiki post #8589 · reply to Tiki post #8588
skamyxatra. > I'm trying to parse out sections of Lojban text delimited by sequences > of NIhO cmavo into their respective paragraphs, sections, chapters, > etc. ... > Does anyone here know how I could use contetx-free grammar rules to > parse the different sections separated by NIhO sequences? > > Any ideas (expressed in EBNF, Prolog, YACC, or whatever you speak) > would be much appreciated! .skamyxatra
If the length of a NIhO sequence exceeds the maximum depth of the parse tree/list/structure, can't you just enclose the list in another list until the depths match? E.g., when you have the list X=[[broda, broda], [broda]], and you encounter four NIhOs in a row, let X=X (two levels of lists because four minus the depth of X is two), and then append to X whatever comes after that.
I don't think this can be handled by a CFG without encoding the lengths of the NIhO strings into the productions, which would lead to an infinitely large grammar. Note that the official Yacc and BNF grammars treat a sequence of NIhOs as a single NIhO and leave the structuring of the text to whatever semantic engine comes after.
mu'omi'e .kamymecraijun.
-- li'a .e'i ca vondei .i mi na'e pu'i kufra loi vondei
To unsubscribe from this list, send mail to [email protected]
with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if
