Evolving Explanatory Novel Patterns for Semantically-Based Text Mining - Natural Language Processing and Text Mining - page 165

scored which is well-correlated with the expert's assessment. Nevertheless, the hy-
pothesis is successful in detecting hidden relations between certain areas ( Mexican
snowed hills ) and the Pinus production and cutting.
9.5 Conclusions
Unlike traditional approaches to Text Mining, in this chapter we contribute an inno-
vative way of combining additional linguistic information and evolutionary learning
techniques in order to produce novel hypotheses which involve explanatory and ef-
fective novel knowledge.
From the experiments and results, it can be noted that the approach supports the
claim that the evolutionary model to KDT indeed is able to find nuggets in textual
information and to provide basic explanations about the hidden relationships in
these discoveries.
We also introduced a unique approach for evaluation which deals with semantic
and Data Mining issues in a high-level way. In this context, the proposed representa-
tion for hypotheses suggests that performing shallow analysis of the documents and
then capturing key rhetorical information may be a good level of processing which
constitutes a trade off between completely deep and keyword-based analysis of text
documents. In addition, the results suggest that the performance of the model in
terms of the correlation with human judgements are slightly better than approaches
using external resources as in [26]. In particular criteria, the model shows a very
good correlation between the system evaluation and the expert assessment of the
hypotheses.
The model deals with the hypothesis production and evaluation in a very promis-
ing way which is shown in the overall results obtained from the experts evaluation
and the individual scores for each hypothesis. However, it is important to note that
unlike the experts who have a lot of experience, preconceived concept models and
complex knowledge in their areas, the system has done relatively well only exploring
the corpus of technical documents and the implicit connections contained in it.
From an evolutionary KDT viewpoint, the correlations and the quality of the
final hypotheses show that the GA operations and the system's evaluation of the
individuals may be effective predictions of really useful novel knowledge from a user
perspective.
References
1. A. Bergstron, P. Jaksetic, and P. Nordin. Acquiring Textual Relations Auto-
matically on the Web Using Genetic Programming. EuroGP 2000, Edinburgh,
Scotland , pages 237-246, April 2000.
2. Michael Berry. Survey of Text Mining: Clustering, Classification, and Retrieval .
Springer, 2004.
3. M. Berthold and D. Hand. Intelligent Data Analysis . Springer, 2000.
4. C. Coello. A Short Tutorial on Evolutionary Multiobjective Optimisation. ACM
Computing Surveys , 2001.