Auto-classification models
Auto-classification models, also called automated document categorization, integrates the IBM® Content Classification's classification model into the IBM® StoredIQ® infoset-generation process.
IBM Content Classification
IBM Content Classification helps organize unstructured content by analyzing the full text of documents and emails and applying rules that automate classification decisions. By managing documents and email, you organize and act on content every day. IBM Content Classification reduces the burden of manual decision making that is done by employees by accurately and automatically organizing information. Embedded with natural language processing and semantic analysis capabilities, IBM Content Classification determines the true intent of words and then uses that knowledge to automate decision making. Unlike other classification systems that are based on rules only, IBM Content Classification combines rules and contextual analysis to incorporate synchronous learning that adapts to changing business needs. As a result, classification becomes accurate over time.
IBM Content Classification can organize information by policies or keywords. The classification process not only searches for a single word or phrase, but also analyzes the entire document. It then distills the main point of the text and assigns the text to a category. When it analyzes content, IBM Content Classification can recognize misspellings, abbreviations, jargon, and technical terms.
Accuracy improves over time because the system adapts to the unique nature of your business by identifying different categories from examples that you provide. You provide feedback, and a deferred retraining operation adjusts the model and implements corrections based on that feedback. The accuracy of the classification results keeps pace with changes in your business.
IBM Content Classification combines this context-based approach with a rule-based, decision-making approach. The system can identify keywords, patterns, such as account numbers and phone numbers of case identifiers, and words within certain proximity of each other. For example, the system can identify occurrences of the phrase "Attorney General" in the same sentence as the word "California". When content that matches a condition in a rule is detected, the action that is defined for the rule is applied, and the document or email is classified.
Examples of classification applications
- Enterprise content standardization. To support document classification and taxonomy automation within your content management system, document properties or metadata can be automatically assigned when the content is classified. Documents can be automatically moved to the correct enterprise repository.
- Compliance and records management. Documents and email can be declared as records when they are classified and placed under the control of record retention policies and compliance standards.
- E-discovery readiness. Documents and email can be filtered to ensure that only items with business value are classified and archived. You can quickly and cost effectively prepare content for potential legal notices.
- Business process optimization. Automated decision making ensures more consistent outcomes and
reduced costs. With content-based analysis, you accomplish these actions.
- Insert documents into the workflow of a business plan.
- Reroute email.
- Suggest and apply agent responses within a customer-support center.
Using auto-classification, you can create and update classification models to help map data within infosets. The Step-up Snippet action extracts and stores text on data objects within an infoset. For auto-classification to work, you must use the Step-up Snippet action since infoset objects without extracted snippets do not have auto-classification scores.
The Data Expert can use IBM Content Classification to train a classification model. The classification model is then registered with IBM StoredIQ Administrator. The registered classification model can be applied to an existing infoset in IBM StoredIQ Data Workbench to generate new metadata for the objects in the infoset. Finally, this metadata can be used in rule-based filters to create new infosets.