Implement n-best output in lt-proc

[ftyers]

Implement an option to lt-proc to output n-best paths. We can use the same option names as in hfst-proc:

  -N N, --analyses=N      Output no more than N analyses
                          (if the transducer is weighted, the N best analyses)
  --weight-classes N      Output no more than N best weight classes
                          (where analyses with equal weight constitute a class

They should work for both analysis and generation.

[Techievena]

@ftyers Please give some directions on how to calculate the weights and start working on this issue.. like the files and classes I should begin modifying first.

[Techievena]

Moreover there is no way we can test this unless we modify the parser to read weights from dix files. Because until the parsers are able to read weights from dix file, it is always the default value of weights we are considering, i.e. 0.0000

[ftyers]

@Techievena you could compile an ATT file with weights in using HFST:

$ cat test.lexc 
Multichar_Symbols

%<n%> %<vblex%> 

%<sg%> %<pl%> %<ger%> %<pres%> %<p3%> %<inf%>

LEXICON Root

Verbs ;
Nouns ;

LEXICON N 

%<n%>%<sg%>: # "weight: 1.0" ;
%<n%>%<pl%>:s # "weight: 1.5" ; 

LEXICON V

%<vblex%>%<inf%>: # "weight: 1.0" ;
%<vblex%>%<pres%>: # "weight: 1.0" ;
%<vblex%>%<pres%>%<p3%>%<sg%>:s # "weight: 1.0" ; 


LEXICON Verbs

walk:walk V "weight: 1.0" ; 

LEXICON Nouns 

walk:walk N "weight: 1.5" ; 

You can see the strings:

$ hfst-lexc test.lexc -o - | hfst-fst2strings -w
hfst-lexc: warning: Defaulting to OpenFst tropical type
Root...2 N...2 V...3 Verbs...1 Nouns...
walk<n><sg>:walk	2.5
walk<n><pl>:walks	3
walk<vblex><pres>:walk	2
walk<vblex><inf>:walk	2
walk<vblex><pres><p3><sg>:walks	2

And print them:

$ hfst-lexc test.lexc -o - | hfst-fst2txt > test.att
hfst-lexc: warning: Defaulting to OpenFst tropical type
Root...2 N...2 V...3 Verbs...1 Nouns...
$ cat test.att 
0	1	w	w	2.000000
1	2	a	a	0.000000
2	3	l	l	0.000000
3	4	k	k	0.000000
4	5	<n>	@0@	0.500000
4	6	<n>	s	1.000000
4	7	<vblex>	@0@	0.000000
4	8	<vblex>	s	0.000000
5	9	<sg>	@0@	0.000000
6	9	<pl>	@0@	0.000000
7	9	<pres>	@0@	0.000000
7	9	<inf>	@0@	0.000000
8	10	<pres>	@0@	0.000000
9	0.000000
10	5	<p3>	@0@	0.000000

And compile:

$ lt-comp lr test.att test.bin
main@standard 11 14

[Techievena]

Thanks a lot for clearing this out for me how to get the att files. This thing must be included in the att_format wiki of Apertium. I have been scratching my head for a long time what you meant by transducer.att files. I will modify the tests of the other PR by tommorow. Thanks a lot!!!