next up previous
Next: Grammar Comparisons Up: On NLP Formalisms: from Previous: Linear Indexed Grammars

Global Indexed Grammars

 

GIGs are based the idea of providing additional descriptive power to handle some canonical NLP problems, where LIGs fail. The canonical NLP problems which exceed context-free power are: multiple agreements, reduplication, and crossing dependencies [4]. We explain the problems in certain details in the following section to compare different formalisms towards NLP. Here, we give the definition of GIGs from [4]:

A GIG is a 6-tuple tex2html_wrap_inline926 where N, T, I are finite pairwise disjoint sets same as the LIG definition. tex2html_wrap_inline930 is the start stack symbol (not in I, N, T) and P is a finite set of productions, having the following form:

The derives relation tex2html_wrap_inline950 on sentential forms are defined as: Let tex2html_wrap_inline952 and tex2html_wrap_inline954 be in tex2html_wrap_inline956, tex2html_wrap_inline958 be in tex2html_wrap_inline960, x in I, w be in tex2html_wrap_inline968 and tex2html_wrap_inline970 in tex2html_wrap_inline972. Then,

  1. If tex2html_wrap_inline974 is a production of type (a.) (i.e., tex2html_wrap_inline976 or tex2html_wrap_inline978) then:
    tex2html_wrap_inline980 or tex2html_wrap_inline982
  2. If tex2html_wrap_inline984 is a production of type (b.) (push), then:
    tex2html_wrap_inline986
  3. If tex2html_wrap_inline974 is a production of type (c.) (pop), then:
    tex2html_wrap_inline990

It has been showed that recognition of the language generated by a GIG is in bounded polynomial time: tex2html_wrap_inline992. An algorithm to construct an LR parsing table for global indexed language is presented in [5], which also shows that it is a full AFL. One thing needs to be pointed out is that global indexed languages may contain languages not in indexed languages (MIX language, Gazdar, 1988). We provide the case in Example 3 [4].

Example 3. (MIX language)

tex2html_wrap_inline994 where P is: tex2html_wrap_inline998, tex2html_wrap_inline1000, tex2html_wrap_inline1002, tex2html_wrap_inline1004, tex2html_wrap_inline1006, tex2html_wrap_inline1008, tex2html_wrap_inline1010, tex2html_wrap_inline1012, tex2html_wrap_inline1014, tex2html_wrap_inline1016, tex2html_wrap_inline1018, tex2html_wrap_inline1020, tex2html_wrap_inline1022

The language generated by tex2html_wrap_inline1024: tex2html_wrap_inline1026 and tex2html_wrap_inline1028.

In comparing with LIGs, GIGs have provided more descriptive power on solving NLP problems. We illustrate this point in next section with examples.


next up previous
Next: Grammar Comparisons Up: On NLP Formalisms: from Previous: Linear Indexed Grammars

Henry Xiao
Fri Apr 29 18:50:42 EDT 2005