The system described below is based on the one described in paper [Per] that uses
directional text relatedness conditions for detecting textual entailment between sentences.
Some other conditions have been experimented with that were generated automatically
using a variant of Genetic Programming. The sentences were searched regardless of the
context in which they appear, as the formulas used were not directly taking that into account.
The word relatedness score required by the formula uses not only identity and synonymy …

There are relatively few entailment heuristics that exploit the directional nature of the
entailment relation. Our system uses directional methods based on the Corley and Mihalcea
formula [3] for expressing the directional relatedness of texts which is then combined with
conditions that must hold for the entailment to be true. The condition used as a starting point
is that of Tatar et al [10]. Several other conditions have been generated automatically based
on the RTE-2009 development dataset using a variant of Genetic Programming. The word …