The Search Engine: TREX

Overview

Search and Classification (TREX) is an integral part of Knowledge Management. It is the technical component that enables the use of search and classification operations to access documents stored in KM repositories. TREX enables:

• Flexible search in large collections of electronic documents
• Mining of interesting and relevant information from documents
• Classification of documents into meaningful structures

TREX can perform these functions on document collections in almost any file format that can be rendered as text, in almost any major language.

Basic Search Functions

TREX offers the following basic functions:

• Exact search for individual words or phrases
  You find only documents that contain exactly the search term or phrase that you entered.
• Boolean search
  You can use logical operators (AND, OR, NOT, NEAR) to build more complex search queries from individual words or phrases.
• Wildcard search
  If you choose not to type a complete search string, you can search using the placeholders '?' and '*' (called wildcard characters or truncation marks) together with a partial string.
• Fuzzy search
  You can search for words or phrases that are spelled like the search phrase, according to an adjustable editing-distance similarity criterion.
• Linguistic search
  You can search for words that represent a linguistic variant of the search query, such as a variant grammatical form.
• Attribute search
  You can search for documents by means of metadata or text attributes (such as document title, author, keywords, or version).

Similarity Search and Classification

TREX offers the following text-mining functions:

• Search for similar terms (related terms)
  If you search for a term, TREX can suggest similar or related terms that you can use to continue your search.
• Search for similar documents (See also … search):
  You can specify a document and search for documents that are similar to that document.
• Determination of key words(feature extraction)
  You can specify a document and receive a list of key words or phrases, also called features that indicate the contents of the document.
• Document classification(taxonomy building)
  You can create your own taxonomy interactively by assigning documents to nodes and adding similar documents to each node, then finding key words for the nodes, using the key words to find further similar documents, and so on.

Knowledge Management API for TREX

The functions of TREX described above are exposed through the Knowledge Management API. The native TREX API is not released for external use.

As indexes are a prerequisite for all search and classification operations, the core interfaces for TREX are related to index management. The IIndexService andIIndex interfaces are used to manage and administer indexes (package com.sapportals.wcm.service.indexmanagement). The actual functionality of an index is accessible through interfaces in sub packages. For example, the package com.sapportals.wcm.service.indexmanagement.retieval.search)  offers search functions. The package com.sapportals.wcm.service.indexmanagement.retieval.classification offers classification functions.

Packages that expose TREX functions are:

• com.sapportals.wcm.service.indexmanagement
• com.sapportals.wcm.service.indexmanagement.retrieval.classification
• com.sapportals.wcm.service.indexmanagement.retrieval.classification.enum
• com.sapportals.wcm.service.indexmanagement.retrieval.search

Further Information

See Integrating 3rd Party Search Engines into KM Index Management to find out how to implement index management to use the search engine of another vendor in Knowledge Management.
See Developing a Search Component to get information on how to integrate a customized search component in a standard KM search iView.