Automatic pattern acquisition for Japanese information extraction

Abstract

One of the central issues for information extraction is the cost of customization from one scenario to another. Research on the automated acquisition of patterns is important for portability and scalability. In this paper, we introduce Tree-Based Pattern representation where a pattern is denoted as a path in the dependency tree of a sentence. We outline the procedure to acquire Tree-Based Patterns in Japanese from un-annotated text. The system extracts the relevant sentences from the training data based on TF/IDF scoring and the common paths in the parse tree of relevant sentences are taken as extracted patterns.

Unsupervised learning of soft patterns for generating definitions from online news

“…To automatically decide whether a sentence is definitional, we could use a simple cutoff in which sentences that are ranked more highly are considered definitional. This is similar to the work by Sudo et al [19] who proposed unsupervised learning method for pattern discovery by utilizing TF-IDF weight to select a set of relevant documents and sentences, and then built patterns from them.…”

“…Similarly, Yangarber et al [21] used a set of basic patterns as "seeds" and learn more scenario oriented extraction patterns automatically. Most relevant to our application of PRF, Sudo et al [19] put forward an unsupervised learning for pattern discovery. They utilized TF×IDF to get a set of relevant documents and sentences and built patterns from them.…”

Comparing information extraction pattern models

“…Several recent approaches to IE have used patterns based on a dependency analysis of the input text (Yangarber, 2003;Sudo et al, 2001;Sudo et al, 2003;Bunescu and Mooney, 2005;. These approaches have used a variety of pattern models (schemes for representing IE patterns based on particular parts of the dependency tree).…”

A local tree alignment-based soft pattern matching approach for information extraction

“…The predicate-argument (SVO) model allows subtrees containing only a verb and its direct subject and object as extraction pattern candidates (Yangarber, 2003). The chain model represents extraction patterns as a chain-shaped path from each target slot value to the root node of the dependency tree (Sudo et al, 2001). A couple of chain model patterns sharing the same verb are linked to each other and construct a linked-chain model pattern (Greenwood and Stevenson, 2006).…”

“…A number of pattern induction approaches have recently been researched based on the dependency analysis (Yangarber, 2003) (Sudo et al, 2001) (Greenwood and Stevenson, 2006) (Sudo et al, 2003). The natural language texts in training instances are parsed by dependency analyzer and converted into dependency trees.…”