Skip to content

Features

Andy Jackson edited this page Mar 2, 2015 · 4 revisions
  • Support both local command-line and Hadoop map-reduce execution.
    • The map-reduce version uses payload hashes to compensate for de-duplication of crawls.
  • Highly configurable.
  • Specific record or content types, or status codes, can be excluded from indexing.
  • Content Analysis:
    • Full-text and metadata extraction from a wide range of formats via Apache Tika.
    • Metadata fields include embedded author information, language detection.
    • Extracts and stores links between resources, at various configurable levels of granularity (URL-to-domain, URL-to-host, URL-to-URL).
    • Extracts embedded licensing information.
    • Stores the ssdeep fuzzy hash of the textual content, allowing documents with similar text to be grouped together.
    • Detects UK postcodes and converts to lat-long coding for Solr, allowing geo-search to be performed.
    • Attempts to use embedded metadata to estimate document creation dates.
    • Uses a simple sentiment analysis algorithm to allow content to be ranked by sentiment.
  • Supports the overlay of additional content annotations (e.g. collections that an item belongs to).
  • Format Analysis:
    • Stores and combines format identification results from Apache Tika and DROID, covering a wide range of formats, versions and encodings.
    • Extracts and stores information on the software tools use to generate the resources.
    • Stores parse errors so problematic format variations can be caught.
    • Stores server content types, file extensions, and header bytes (both the first four bytes, and header byte shingles), allowing previously unidentified formats to be identified.
    • For HTML, can record the elements employed by each resource, allowing element usage to be analysed over time.
    • For PDF, can run each one through Apache Prefight in order to diagnose possible preservation risks.
    • For XML, stores the root element namespace.

Note that many of these features are brand new and in the process of being researched. In many cases, the quality and utility of the results they yield is still to be ascertained.

You can’t perform that action at this time.