34 / 36
Student corner
Working with Islamic domain to extract information and discover any
inconsistency is a very challenging research. The domain is sensitive and risky
because of the important and critical reliance on these data that exist in the
web.
The attempt is done by avoiding deep linguistic analysis for discovering
named entities. Instead, classification approach was used to label texts into
different targeted categories as defined in the initial domain ontology. For
each categorization level, discriminative keywords were extracted using the
hyper-rectangular decomposition method. Then, obtained keywords were fed
into the random forest classifier, which automatically detects the category of
each advisory opinion. Here, in the case of Fatwas, all labels are considered as
premises except judgements considered as consequents.
The results are interesting and show that the proposed method is
promising in the field. The proposed method is applicable for a general
domain. However, it requires the definition of an ontology, the discrimination
of labels into premises and consequents, and contradictory labels in general.
Information extraction and detecting inconsistency
Dr. Jameela Al-Otaibi
( CSE Alumni)
34