Package com.norconex.importer.handler.transformer.impl

Class TitleGeneratorTransformerConfig

java.lang.Object
com.norconex.importer.handler.BaseLabelableConfig
com.norconex.importer.handler.BaseDocHandlerConfig
com.norconex.importer.handler.transformer.impl.TitleGeneratorTransformerConfig
All Implemented Interfaces:
Labelable, ChunkedTextSupport
public class TitleGeneratorTransformerConfig extends BaseDocHandlerConfig implements ChunkedTextSupport

Attempts to generate a title from the document content (default) or a specified metadata field. It does not consider the document format/structure nor does it weight some terms more than others. For instance, it would not consider text found in <H1> tags more importantly than other text in HTML documents.

If isDetectHeading() returns true, this handler will check if the content starts with a stand-alone, single-sentence line (which is assumed to be the actual title). That is, a line of text with only one sentence in it, followed by one or more new line characters. To help eliminate cases where such sentence are inappropriate, you can specify a minimum and maximum number of characters that first line should have with setDetectHeadingMinLength(int) and setDetectHeadingMaxLength(int) (e.g. to ignore "Page 1" text and the like).

Unless a target field name is provided, the default field name where the title will be stored is document.generatedTitle.

Storing values in an existing field

If a target field with the same name already exists for a document, values will be added to the end of the existing value list. It is possible to change this default behavior by supplying a PropertySetter.

If it cannot generate a title, it will fall-back to retrieving the first sentence from the text.

The generated title length is limited to 150 characters by default. You can change that limit by using setTitleMaxLength(int). Text larger than the max limit will be truncated and three dots will be added in square brackets ([...]). To remove the limit, use -1 (or constant UNLIMITED_TITLE_LENGTH).

This class should be used as a post-parsing handler only (or otherwise on unformatted text).

The algorithm to detect titles is quite basic. It uses a generic statistics-based approach to weight each sentences up to a certain amount, and simply returns the sentence with the highest attributed weight given a minimum threshold has been met. You are strongly encouraged to use a more sophisticated summarization engine if you want more accurate titles generated.

Max read size

This tagger will only analyze up to the first 10,000 characters. You can change this maximum with setMaxReadSize(int). Given this class is not optimized for large content analysis, setting a huge maximum number of characters could cause serious performance issues on large large files.

See Also: