Changelog

Version 1.2.11 - 2019-12-12

  • Add an ml_threshold parameter to control machine learning results during classification. Setting this value lower will result in higher recall, but lower precision.
  • async is dropped as an option during classification calls. All classification calls will now return results in the body of the response object rather than allowing for retrieval later.

Version 1.2.10 - 2019-08-14

  • Cache the HTML from URL classification to prevent recrawling the same URL if it is seen within a certain period of time. You can force a recrawl by passing in cache_skip: true in your POST data.

Version 1.2.9 - 2019-07-01

  • Bug fixes
  • Performance enhancements

Version 1.2.8 - 2019-04-10

  • Bug fix for URL classification - all non-200 responses for individually submitted URLs to the API will return a 400.
  • Specify a min_tags parameter in classify/url to fallback to a different extraction engine if not enough content is retrieved
  • Expose a user/allotments call to show usage limits if applicable
  • Bug fix where sometimes overlay output format was incorrect

Version 1.2.7 - 2019-03-05

  • Introduction of Facets to Category objects. Facets can be used to help identify things like brands and products and makes our Categories a little bit richer.

Version 1.2.6 - 2019-01-08

  • Significant changes to the eContext sentiment models as well as significant performance enhancement. Quickly retrieve sentiment on classification calls.
  • New NLP features - language identification for URLs and other input text. Available now at /v2/nlp/lid

Version 1.2.5 - 2018-10-25

  • Added Sentiment to most classification calls. Generally, simply add a "sentiment": true parameter to the request body.
  • Added NLP parsing functions to the API at /v2/nlp/parse which performs POS tagging, NER, dependency parsing, entity extraction, etc. Parsing is currently limited to select users. Please contact eContext.ai if you’re interested in accessing these functions.

Version 1.2.0 - 2018-05-10

  • Include new multi-lingual support language-support
  • Improved the efficiency for Custom Taxonomies
  • Increased performance of the classification engine to provide greater detail and insight into classifications

Version 1.1.6 - 2017-03-16

  • Several performance enhancements and bug fixes
  • Changed error handling on classify/url to be a bit more robust
  • Add client-controlled Custom Taxonomies that can be used by the classification engine in lieu of the default eContext taxonomy

Version 1.1.5 - 2017-02-02

  • Specify a taxonomy_timestamp parameter to freeze the eContext Taxonomy for long-running classification tasks
  • Return more useful errors for invalid filters in keywords/search calls

Version 1.1.4 - 2017-01-16

  • Added scored_keywords entries for items classified in ML Classification results
  • Fixed bug where classify/url would fail if there was a new-line character in the input
  • Significant expansion and updates to internal keyword models

Version 1.1.3 - 2016-11-11

  • Better recognition of duplicate entity categories, particularly names in ML Classification results

Version 1.1.2 - 2016-09-29

  • Added usage limits to assist users in keeping to API usage quotas
  • Integrated entity categories into the main categories object in response objects

Version 1.1.1 - 2016-08-16

  • Specify a branches parameter when classifying keywords to restrict to a particular vertical
  • Specify a best_match parameter (defaults to true) to classify/social, classify/html, classify/url, and classify/text calls which will force the engine to retain submatches instead of removing them from results
  • Added the ability to create custom taxonomies that users may use for specific classification tasks

Version 1.1.0 - 2016-07-11

  • Added fallback NLP Entity extraction options in cases where the eContext Taxonomy does not provide coverage
  • Added keyword flagging to identify keywords that may be sensitive to particular audiences
  • Added social content flagging to identify social media posts that may be sensitive to particular audiences
  • Changed classify/url to return a 403 Status Code if a requested URL is blocked by that domains robots.txt

Version 1.0.24

  • New eContext translation service provides fast translation for incoming content. Supports Spanish, French, Portuguese, Italian, Dutch, German content
  • Translated content is now passed through to your classification results

Version 1.0.23

  • Added the Twitter Interest Taxonomy overlay as an available client add-on
  • Changed format of overlays to be more descriptive - each taxonomy map now returns as a list with Tier 1 and Tier 2, when available

Version 1.0.22

  • Internal improvements

Version 1.0.21

  • Added client taxonomy overlay capabilities
  • Added IAB Taxonomy overlay as an available client addon

Version 1.0.20

  • Improved performance of demo limits for new accounts
  • Internal improvements in dataset migration and publication including

Version 1.0.19

  • Added an “async” parameter to classify/ calls to block on classification. The result of the POST will be the actual classification results rather than a link to the result URI.
  • Reject classify/url POSTs where the url being classified doesn’t provide an apporpriate content-type (‘text/html’, ‘text/xhtml’, ‘application/xhtml+xml’, ‘text/xml’, ‘application/xml’)
  • Reject classify/url POSTs where the url being classified is too large (content-length >= 256000 bytes)
  • Fixed an issue with classify/* results being lost occasionally

Version 1.0.18

  • Added an eContext Plugin to Google Translate API
  • Added a check to avoid translation of content if it is determined to be in English

Version 1.0.17

  • Added /categories/tiers to show all top-tier categories in the eContext Taxonomy
  • Added ability to perform automatic content translation prior to classification using a bring-your-own translation service - currently only Microsoft Translator API
  • Added general Category statistics including Social IDF (Inverse Document Frequency) and Social Relevance
  • Deprecated /classify/twitter - these calls should be handled by /classify/social