The pattern-based extraction methods can be grouped into three categories based on the rules they use or learn. The categories correspond to emph{Dictionary-based methods}, emph{Rule-based methods} and methods based on emph{Wrapper Induction}. Dictionary-based methods aim to automatically learn the ‘dictionary’ of patterns that can be used to identify relevant information in documents. Rule-based methods use several linguistic rules instead of a dictionary to extract relevant information from text. These syntactic or semantic constraints have to be learned together with delimiters that bound the target text.

Rule based OBIE systems use two rule learning algorithms: a emph{bottom-up algorithm} which learns rules from special cases to general ones, and a emph{top-down algorithm} which learns rules the other way around. The methods of this category have mostly been used for semi-structured web documents. The third category consists of methods based on wrapper induction, which is a subtype of rule based methods aimed at structured and semi-structured documents. Wrappers are extraction procedures, which consist of a set extraction rules and program codes that are required to apply these rules. They are used when there exists a repeatable structure to extract information from. Wrapper induction is a technique for automatically learning the wrappers.

Given a training data set, the induction algorithm learns a wrapper for extracting the target information. paragraph{ extbf{Related work in OBIE systems}}Embley’s OBIE systems were the first systems in the field to use pattern-based extraction cite{first}. These systems adopted a wrapper induction approach and combined linguistic rules in the form of regular expressions with the elements of ontologies, resulting in so-called ’emph{extraction ontologies}’ cite{embley}. Embley describes an extraction ontology as an augmented conceptual model that serves as a wrapper for a specific domain of interest. When an extraction ontology is applied to a document, the ontology identifies and extracts objects and relationships associated with named object sets and relationship sets from the ontology’s conceptual model. Yildiz and Miksch cite{ontox} proposed a rule-based approach in emph{OntoX}. OntoX contains a rule generating module that formulates rules as regular expressions based on the knowledge in an OWL input ontology. An extraction module applies these rules to the input texts to determine the most accurate candidate values among them.

This module finally returns the extracted values and makes suggestions to the user regarding possible improvements in the ontology, such as the removal of out-of-date constructs.