Release Notes

Any feature using TensorFlow will not work with Elasticsearch 8.1.1 including:

New language support: Tokenization is now supported for Indonesian, Standard Malay, and Tagalog. POS tags are not supported for these languages, so the universalPosTags option is ignored. (ETROG-3443, ETROG-3465)

Fragment definition: A single line followed by an empty line is no longer always considered a fragment. They are still considered fragments if the line is short, as specified by the maxTokensForShortLine parameter. (ETROG-3431)

TensorFlow supported: This plugin supports TensorFlow. (ESPI-168)

If you will be using any of features which require TensorFlow:

Features that require TensorFlow include:

Improved Hebrew tokenizer and new analyzer: The Hebrew tokenizer is now more consistent with the tokenizers of other languages. Hebrew tokenization and analysis are now done in separate steps.(ETROG-3290)

Normalized characters: Normalized half-width and full-width characters are processed the same as their counterparts. (ETROG-3351)

Improved directory structure: The contents of the models/ directory are now separated into subdirectories by language. (ETROG-1218)

Statistical models moved to models/ directory: The following files have been moved from dicts/ to models/: (ETROG-1218)

New option for tokenizers: We've added a new option, tokenizerType to specify which tokenizer to use. The options alternativeTokenization and fstTokenize are deprecated in favor of tokenizerType. (ETROG-3419)

New Korean tokenizer: We've added a new tokenizer for spaceless Korean input. The previous tokenizer was not trained on spaceless Korean and did not perform well without spaces between tokens. Activate it by setting tokenizerType to spaceless_statistical. (ETROG-3392)

Hebrew tokens containing a geresh are now tokenized properly. Previously, only the part up to the geresh would be returned as the token text, and the part after the geresh would sometimes be considered a suffix. Now the whole token is returned as the token's text. (ETROG-3262, ETROG-3290)

Example: מע'רב

Multiple punctuation characters are no longer returned as a single token in Chinese when alternativeTokenization is true or tokenizerType is set to spaceless_lexical. Now each character is its own token. (ETROG-3402)

Example: Input: 天津？？

The tools directory is now included in the RBL-Elasticsearch package. (ESPI-31)

Greek lexicon: The Greek lexicon has additional words. (ETROG-3288)

Greek disambiguation improved: Certain Greek forms are now disambiguated to prefer a modern analysis over an archaic analysis. alternativeGreekDisambiguation must be set to false, which is the default. (ETROG-3289)

Example: δείξε

New Greek disambiguator added: The new Greek disambiguator is more accurate, but slower. The new disambiguator is enabled by default. To use the old disambiguator, set alternativeGreekDisambiguation to true. (ETROG-3304)

RBL-JE no longer normalizes certain emoji ZWJ sequences to U+1F48F KISS, U+1F491 COUPLE WITH HEART, and U+1F46A FAMILY, to be consistent with Unicode’s efforts to make emoji more gender-neutral by default. (ETROG-3350)

The Greek guesser now handles tokens with non-alphanumeric characters. (ETROG-3286)

Example: Start+

Performance improvement: Spanish disambiguation with alternativeSpanishDisambiguation set to false is now faster. (ETROG-3271)

Performance improvement: Korean disambiguation is now faster. (ETROG-3280, ETROG-3282)

Support for unknown language: If the language is unknown (xxx), tokenization and sentence breaking is supported. (ETROG-3278)

Solr 8.7.0: We now support Solr 8.7.0. (ETROG-3315)

Tokenization rule preprocessor: The preprocessor command !!btinclude is supported in tokenization rule files, supporting inclusion of files in rule files. (ETROG-2497)

Updated sample: The tokenize-analyze sample has been changed from two applications running in sequence to a single application that both tokenizes and analyzes. (ETROG-3291)

New sample: The sample csc-annotate demonstrates using CSC with the ADM API. (ETROG-3317)

Deprecated option: TokenizerOption#includeRoots has been deprecated and replaced with TokenizerOption#includeHebrewRoots. (ETROG-3314)

Deprecated option: The alternative tokenization option deliverExtendedAttributes is now deprecated. Previously it delivered an unsupported extended property. (ETROG-3311)

Combining characters in Hebrew which were being erroneously split into tokens separate from their bases are now not being split. (ETROG-3277)

A clear exception (RosetteUnsupportedLanguageException) is now thrown when tokenizing some unsupported languages. Previously, these languages appeared to work. The same tokenizer is still available by specifying the unknown language (xxx). The languages impacted are Albanian, Bulgarian, Croatian, Indonesian, Malay, Slovenian, Standard Malay, and Ukrainian. (ETROG-3278, ETROG-3326)

RBL no longer crashes when alternativeTokenization and fragmentBoundaryDetection are both enabled for some inputs in Japanese and Chinese. (ETROG-3285)

Correct start and end offsets are now produced when fstTokenize is set to true. Previously, some Spanish inputs would produce tokens with start and end offsets of 0. (ETROG-3292)

The mappings of default Basis POS tags to universal POS tags (UPT-16) have been corrected for Greek. (ETROG-3306)

Tokens no longer have null token types. (ETROG-3316)

When an NFKC normalized character results in multiple tokens, those tokens no longer have equal start and end offsets. Previously this could occur when nfkcNormalize was set to true. (ETROG-2505)

Example: ﷺ