A Case Study in Natural Language Based Web Search - Natural Language Processing and Text Mining - page 72

For instance, the parse tree for our Washington example above is shown in Table

5.1.

Table 5.1. The parse tree representation of a sentence.

Head

Label Modifier Root POS Label Role Antecedent Attributes

1 Appointed Appoint V

2 commander N 1 Obj Person/title

3 of Prep 2 Mod

4 the Det 5 Det

Continental

5 N 3 Pcomp Organization/name

Army

6 in Prep 1 Mod

7 1775 N 6 Pcomp Numeric/date

George

8 N 9 Subj Person/name

Washington

9 molded mold V

10 a Det 12 Det

11 fighting A 12 Mod

12 force N 9 Obj

13 that N 15 Subj 12

14 eventually A 15 Mod

15 won win V

16 independence N 15 Obj

17 from Prep 16 Mod

Great

18 N 17 Pcomp Location/country

Britain

Fig. 5.2. The Subject-Action-Object indexing structure.

The basic idea behind or approach to indexing involves collapsing selected nodes
in the parse tree to reduce the overall complexity of the dependency structures.