Toolbox

Classifier management

A classifier is a tool that supports the classification and routing of documents to multiple templates and models. By defining keywords or providing training samples, you can establish classification criteria to automatically match documents with published custom templates or deployed information extraction models. After you publish a classifier, you can use a single classifier API to perform structured recognition and information extraction for various types of business data. This eliminates the need to manually sort data before calling individual template or model APIs. You can also adjust a classifier's categories by republishing it.

The classifier details page includes Basic Information (Classifier Name, classifier ID, Number of Categories), Training Status (click the Train button to start training), and a Category Management section. The category management table displays the Category, Name, Identifier, Classifier Keyword, and Number of Training Samples for each category. For each category, the available actions are Edit Keywords, Edit Training Set, and Delete.

Keyword: A unique piece of text found within a specific data category. It is used to match a template or model to an input image. You must select a keyword that appears only in the current template or model. This allows you to use the keyword to uniquely determine the category to which an image belongs. For example, when creating a classifier for ID cards and household registration books, you can select keywords such as "Education Level", "Employer", or "Certificate Number" for recognizing household registration books. For ID cards, you can only select "Citizen Identity Number" as the keyword.

Training set: For template-based tasks, you can improve classification accuracy by uploading a compressed file of images that share the same layout. For templates with complex layouts, we recommend uploading at least 20 unique images of the same layout.

Try out: After a classifier is trained, you can test its classification performance.

Field type management

A field type: A feature that allows you to incorporate dictionaries of general, business, or industry knowledge into the field recognition and extraction process on the self-learning platform. It uses field correction and format normalization, which improves field recognition accuracy and standardizes field output.

• General-purpose field: A system-predefined field type with common attributes.

• Custom field: A unique field type that you can create based on your business requirements. You can create a custom field by creating a dictionary of enumerated values, such as national administrative divisions, lists of candidate names, or days of the week.