Next: Bibliography
Up: Evaluation and Results
Previous: Comparison with Alvey
The evaluation in this section was done with the earlier 1995 release
of the grammar. This section compares the performance of XTAG against
that of the CLARE-2 system ([#!clare-report92!#]) on the ATIS corpus.
Table G.8 shows the performance results. The
percentage parsed column for both systems represents the percentage of
sentences that produced any parse. It must be noted that the
performance result shown for CLARE-2 is without any tuning of the
grammar for the ATIS domain. The performance of CLARE-3, a later
version of the CLARE system, is estimated to be 10% higher than that
of the CLARE-2 system.31.6
System |
Mean length |
% parsed |
CLARE-2 |
6.53 |
68.50% |
XTAG |
7.62 |
88.35% |
- {Performance of CLARE-2 and XTAG on the ATIS corpus
In an attempt to compare the performance of the two systems on a wider
range of sentences (from similar genres), we provide in
Table G.9 the performance of CLARE-2 on LOB corpus
and the performance of XTAG on the WSJ corpus. The performance was
measured on sentences of up to 10 words for both systems.
System |
Corpus |
Mean length |
% parsed |
CLARE-2 |
LOB |
5.95 |
53.40% |
XTAG |
WSJ |
6.00 |
55.58% |
- {Performance of CLARE-2 and XTAG on LOB and WSJ corpus respectively
Next: Bibliography
Up: Evaluation and Results
Previous: Comparison with Alvey
XTAG Project
http://www.cis.upenn.edu/~xtag