Data Mining - Mehmed Kantardzic [181]
text refining, which transforms free-form text documents into a chosen intermediate form (IF), and
knowledge distillation, which deduces patterns or knowledge from an IF.
Figure 11.6. A text-mining framework.
An IF can be semi-structured, such as a conceptual-graph representation, or structured, such as a relational-data representation. IFs with varying degrees of complexity are suitable for different mining purposes. They can be classified as document-based, wherein each entity represents a document, or concept-based, wherein each entity represents an object or concept of interests in a specific domain. Mining a document-based IF deduces patterns and relationships across documents. Document clustering, visualization, and categorization are examples of mining from document-based IFs.
For a fine-grained, domain-specific, knowledge-discovery task, it is necessary to perform a semantic analysis and derive a sufficiently rich representation to capture the relationship between objects or concepts described in the document. Mining a concept-based IF derives patterns and relationships across objects and concepts. These semantic-analysis methods are computationally expensive, and it is a challenge to make them more efficient and scalable for very large text corpora. Text-mining operations such as predictive modeling and association discovery fall in this category. A document-based IF can be transformed into a concept-based IF by realigning or extracting the relevant information according to the objects of interests in a specific domain. It follows that a document-based IF is usually domain-independent and a concept-based is a domain-dependent representation.
Text-refining and knowledge-distillation functions as well as the IF adopted are the basis for classifying different text-mining tools and their corresponding techniques. One group of techniques, and some recently available commercial products, focuses on document organization, visualization, and navigation. Another group focuses on text-analysis functions, IR, categorization, and summarization.
An important and large subclass of these text-mining tools and techniques is based on document visualization. The general approach here is to organize documents based on their similarities and present the groups or clusters of the documents as 2-D or 3-D graphics. IBM’s Intelligent Miner and SAS Enterprise Miner are probably the most comprehensive text-mining products today. They offer a set of text-analysis tools that include tools for feature-extraction, clustering, summarization, and categorization; they also incorporate a text search engine. More examples of text-mining tools are given in Appendix A.
Domain knowledge, not used and analyzed by any currently available text-mining tool, could play an important role in the text-mining process. Specifically, domain knowledge can be used as early as in the text-refining stage to improve parsing efficiency and derive a more compact IF. Domain knowledge could also play a part in knowledge distillation to improve learning efficiency. All these ideas are still in their infancy, and we expect that the next generation of text-mining techniques and tools will improve the quality of information and knowledge discovery from text.
11.7 LATENT SEMANTIC ANALYSIS (LSA)
LSA is a method that was originally developed to improve the accuracy and effectiveness of IR techniques by focusing on semantic meaning of words across a series of usage contexts,