Releases: terrier-org/terrier-core
5.11
What's Changed
Minor update - some improvements for PyTerrier's IterDictIndexer; and some efficiency improvements, particularly for multi-threaded retrieval.
Indexing
- allow indexing fields to be separated from FlatJSONDocument parsing by @cmacdonald in #257
Retrieval
- Remove controlNonVisibleCharacters by @cmacdonald in #260
- Improvements for concurrent indices by @cmacdonald in #259
Minor:
- Bump most JDK version tested in GHA to 21 by @cmacdonald in #254
Full Changelog: 5.10...5.11
5.10
Significant update, adding functionality useful for PyTerrier use cases, and some other small minor improvements.
Retrieval
- Include RM1 and RM3 query expansion models, originally authored by @ntonellotto, and previously included in the separate terrier-prf package. This integration aims to simplify the usage in PyTerrier #250, #253. Thanks to @mam10eks for RM3 improvements.
- Include previously unreleased doc-vectors module authored by @cmacdonald #249, which allows calculation of additional weighting model features using direct index rather than the inverted index access provided by Fat. For more information, see the relevant README. This functionality can be accessed through PyTerrier for the purposes of learning-to-rank pipelines by using
FeaturesBatchRetrieve(..., method='dv')
.
Other
- Bump slf4j and logback versions due to logging conflict with Anserini in PyTerrier #251.
- suppress "No etc/terrier.properties message" that always appears at PyTerrier startup #252.
Full Changelog: 5.9...5.10
5.9
Minor release, mostly addressing #246
What's Changed
- Prevent negative free memory from ending indexing by @cmacdonald in #245
- Fix problem with FatFull misassociating postings to documents by @cmacdonald in #246
- Bump ch.qos.logback:logback-classic from 1.2.0 to 1.2.13 in /modules/logging by @dependabot in #244
- Bump ch.qos.logback:logback-classic from 1.2.0 to 1.2.13 in /modules/rest-server by @dependabot in #243
Full Changelog: 5.8...5.9
5.8
What's Changed
- addresses #215 about FLATJsonDocument error handling by @cmacdonald in #216
- Bump jackson-databind from 2.13.4.1 to 2.13.4.2 in /modules/core by @dependabot in #220
- [SECURITY] Fix Temporary File Information Disclosure Vulnerability
by @JLLeitschuh in #213 - make TermPipelines IndexConfigurable at querying time by @cmacdonald in #218
- Fix weight propagation for ExplicitMultiTermQuery by @cmacdonald in #229
- addresses issue #225 - MultiIndex does not expose index structures inputstreams by @cmacdonald in #226
- Bump guava from 29.0-jre to 32.0.0-jre in /modules/core by @dependabot in #224
- bump GHA versions by @cmacdonald in #231
- emit warning for fielded indices by @cmacdonald in #232
- include a shaded version of snowball in Terrier by @cmacdonald in #230
New Contributors
- @monika-1315 made their first contribution in #196
- @JLLeitschuh made their first contribution in #213
Full Changelog: 5.6...5.8
5.6
Terrier 5.6 - 17/09/2021
Minor update, making configuration from PyTerrier easier, particularly use of the Terrier Data Repository, and addressing small inconsistencies. Note that the method of configuring weighting model paraleters has now changed - see the relevant documentation.
Retrieval
- #182 QueryParserException message should give parser's Exception message (with thanks to Eugene Yang and Sean MacAvaney)
- #177 #185 use stable sort when sorting result set; make DAAT retrieval stable too (with thanks to Sean Macavaney and Nicola Tonellotto)
- #179 models configuration from controls
- #122 #181 Determine termpipeline configuration (stemming etc) from the index
Other
5.5
Significant update, introducing alternative compression for the MetaIndex, reentrant Matching and data structure classes, and addressing bugs and enhancements identified through usage from PyTerrier. From 5.5 onwards, the default compression of the MetaIndex will be Zstd, which achieves faster decompression with comparable compression rates. Existing MetaIndexs can still be read.
See the Whats New documentation for more information on this release.
5.4
Minor update addressing bugs identified through usage from PyTerrier and others.
5.3
Significant update, addressing various capabilities that can be utilised by PyTerrier -- the new Python bindings for Terrier -- as well as addressing other feedback from users. This release continues to reduce the reliance on ApplicationSetup. In particular, for retrieval, more configuration is available through controls. We have also trimmed out older packages that are no longer used (c.f. integer-compression and website-search) into separate legacy Terrier components. Terrier has enhanced support for loading external components from local and remove Maven repositories, including Jitpack. Components compatible with this release include ones for reading Lucene indices, reading CIFF files as a Terrier index, as well as additional query expansion models (incl. RM3 and Axiomatic query expansion).
5.2
5.2
5.1
5.1 release