发布于 2016-02-16 00:24:34 | 188 次阅读 | 评论: 0 | 来源: 网友投递
Apache Tika 内容抽取工具集合
Apache Tika 利用现有的解析类库，从不同格式的文档中（例如HTML, PDF, Doc)，侦测和提取出元数据和结构化内容。
Apache Tika 1.12 发布，Tika是一个内容抽取的工具集合(a toolkit for text extracting)。它集成了POI, Pdfbox 并且为文本抽取工作提供了一个统一的界面。其次，Tika也提供了便利的扩展API，用来丰富其对第三方文件格式的支持。
该版本包含不少改进和 bug 修复。具体内容包括：
* Slide notes are now linked to the slide XHTML in the PPT output
(TIKA-1840).
* JSON tests in Tika server were updated to remove impossible casts
(Github-73).
* Fix bug in GeoTopicParser where NER is reused instead of instantiated
with each request (TIKA-1834).
* Upgrade rome to 1.5.1 && Downgrade Rome dependency to 0.9 to avoid
nasty NPE (TIKA-1820, TIKA-1516)
* The NamedEntityParser was enhanced to generate text content
in addition to metadata (TIKA-1815, TIKA-1816).
* A significant speed-up is made to the GeoTopicParser by
using the new REST server capabilities from Lucene Geo
Gazetteer (TIKA-1803).
* A parser to compute motion properties in Videos, e.g.,
Histogram of Oriented Gradients and Histogram of Optical Flows
using the Pooled Time Series algorithm, was added (TIKA-1798).
* Provide NamedEntityParser which exposes Named Entity Recognition
from OpenNLP and Stanford NER providers (TIKA-1787, GitHub-61,
GitHub-62).
* Allow XHTMLContentHandler to pass attributes of html element
via Markus Jelsma (TIKA-1782).
* Fix regression with spacing in PPT via Andreas Beeker (TIKA-1777).
* Tika Facade parse methods for Path and File added which take a
Metadata object, to mirror the existing InputStream one (GitHub-60)
* GeoParser fix for loading the NER model from a jar file (TIKA-1791)