版本说明

本页面记录了 Natural Language 的正式版更新。我们建议 Natural Language 开发者定期查看此列表,以便及时获取任何新公告。对于重大更新,我们还将通过 cloud-nl-discuss 邮寄名单发布公告。

您可以在 Google Cloud 版本说明页面上查看 Google Cloud 所有产品的最新产品动态。

如需接收最新产品动态,请将本页面的网址添加到您的 Feed 阅读器,或直接添加 Feed 网址:https://cloud.google.com/feeds/language-release-notes.xml

August 28, 2023

The Natural Language API v2 is now available in Public Preview. This new version of the API supports Sentiment Analysis, Entity Analysis, Text Moderation, and Content Classification. This version has major updates for Sentiment Analysis and Entity Analysis, which include performance upgrades and general improvements. All API calls now return language_code .

Sentiment analysis uses a new PaLM-based model for improved quality. The language field is now called language_code.

Entity analysis uses a new PaLM-based model for improved quality. Please note the following updates:

  • language field is now called "language_code"
  • No salience field.
  • No wikipedia_url as metadata.
  • Parsing improvements for the following entity types: PHONE_NUMBER, ADDRESS, DATE, NUMBER, PRICE
  • New probability score field is returned for all entities where NUMBER, PHONE_NUMBER, ADDRESS, PRICE, DATE will always be 1.0.

January 31, 2023

Natural Language Content Classification v2 model is now Generally Available. This model supports an expanded taxonomy with 1091 content categories and 11 languages. The model is distilled from a Large Language Model with improved performance over the v1 offering.

October 18, 2022

Natural Language Content Classification v2 model is now in Public Preview. This model supports an expanded taxonomy with 1091 content categories and 11 languages. The model is distilled from a Large Language Model with improved performance over the v1 offering.

March 20, 2020

The Natural Language API now supports additional languages for sentiment analysis.

February 20, 2020

The Natural Language API now offers two multiregion endpoints (us-language.googleapis.com and eu-language.googleapis.com). Using a multiregion endpoint enables you to configure the Natural Language API to store and perform machine learning on your data in the United States or European Union.

May 03, 2019

The Natural Language API now supports Spanish for entity sentiment analysis.

April 04, 2019

The Natural Language API now supports Russian for entity analysis and syntactic analysis, and Japanese for entity sentiment analysis.

The analyzeEntities method identifies and returns additional entity types: phone numbers, addresses, dates, prices, and numbers.

November 16, 2017

Cloud Natural Language API version 1.2 GA release.

The Natural Language API has released the new classifyText method that analyzes text content and returns a content category for the content. For more information, see Content Classification.

September 19, 2017

The Natural Language API v1 now supports entity sentiment analysis with the analyzeEntitySentiment method. Entity sentiment analysis is currently only available for the English language. For more information, see Entity Sentiment Analysis.

The Natural Language API v1beta2 has added beta support for a new classifyText method that analyzes text content and returns a content category for the content. For more information, see Content Classification.

August 04, 2017

The following languages are now fully supported for entity, sentiment, and syntax analysis: * Chinese (Simplified and Traditional) * French * German * Italian * Korean * Portuguese

Sentiment analysis has been updated. You might see minor differences in score and magnitude from the same request made to previous releases.

Entity analysis has been updated. You might see minor differences in metadata (Wikipedia URLs, knowledge graph MIDs) of entities identified in a block of text from the same request made to the previous releases.

May 02, 2017

Cloud Natural Language API v1beta2 release. Added beta support for entity, sentiment, and syntax analysis for the following languages: * Chinese (Simplified and Traditional) * French * German * Italian * Korean * Portuguese

Added beta support for entity sentiment analysis (analyzeEntitySentiment). Entity sentiment analysis is currently only available for the English language. For more information, see Entity Sentiment Analysis.

The v1beta2 release includes updated document sentiment analysis. You might see differing results from the same request made to the Natural Language API v1 and v1beta2 releases.

The v1beta2 release includes updated entity analysis. You might see minor differences in metadata (Wikipedia URLs, knowledge graph MIDs) from the same request made to the Natural Language API v1 and v1beta2 releases.

The Natural Language API v1beta1 release has been deprecated.

November 15, 2016

Promoted the Cloud Natural Language API from Beta (v1beta1) to General Availability (v1). Updates for v1 from the previous v1beta1 API endpoint include the features listed below.

The sentiment analysis feature has these enhancements: * Support for Japanese (ja) and Spanish (es) * Availability of sentence level sentiment * Support for EncodingType within a Sentiment Analysis request to calculate offsets for sentence-level sentiment *Addition of a score field within the Sentiment type to replace the previous polarity field

The entity analysis feature has these enhancements: * Added Google Knowledge Graph MID values, if available, to the metadata field for each Entity returned * Returns Entity Mention Types for both proper nouns and common nouns (known as "nominals")

Syntactic analysis has these enhancements: * Added a new analyzeSyntax method * Added morphology support to returned tokens

Removed the polarity field within the Sentiment type in favor of the new score field.