[
Previous: DFR Description] [
Contents] [
Next: Bibliography]
Terrier Future Features and Known Issues
List of features and known issues that are marked for future Terrier versions:
Future Features
- New types of indices. In Terrier 4.0, the central Index class will become abstract and will be extended by different types of index. Current index functionality will move to new type of index called IndexOnDisk.
- The structures within an index will become more generic. In Terrier 4.0, Inverted and Direct indices will become PostingIndex instances instead of being separate classes like InvertedIndex. New code should use PostingIndex where possible.
- New index types that can be updated incrementally will be added In Terrier 4.0. These will allow new documents to be added on-the-fly for real-time search applications.
- Functionality to create memory-only indices that can be incrementally updated over time.
- Functionality to create, modify and search over groups of separate indices as if they were one index.
- Improved Learning-to-Rank support with result sets that can hold document feature information.
- Move towards a modular sub-package structure to facilitate the release of non-core functionalities.
- The addition of Terrier's own Exceptions for setup, indexing, querying.
Known Issues
- Query language should be refined: In particular, ambiguity warnings when generating the parser with ANTLR should be removed, and a more structured query language should be introduced.
Deprecations, Scheduled Refactorings
The following classes and interfaces are/may be deprecated in this version of Terrier and will likely be removed or refactored in a future release:
- Indexer and org.terrier.structures.indexing hierarchy may be refactored in a future release.
- ApplicationSetup will be replaced with a new non-global configuration API.
- The five field weighting model score() methods are being deprecated since only the two field versions are used.
- Terrier currently uses Hadoop version 0.20. Future releases of Terrier will upgrade to a later branch. Instructions to upgrade Terrier yourself can be found here
Contributions
All community contributions to the Terrier framework are welcome.
In addition, you can find more information about contributing on
the Terrier website.
[
Previous: DFR Description] [
Contents] [
Next: Bibliography]