ChangeLog
Details of changes made to the Haven OnDemand platform and APIs.

Change Log for Haven OnDemand


3 April 2017

This section describes the enhancements and resolved issues in the April 2017 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Face Detection API has been added. This API finds faces in media files.

  • The Image Classification API has been added. This API classifies images in media files.

  • The Object Recognition API has been added. This API matches logos in media files.

  • The Scene Change Detection V2 API has been added which simplifies the response.

  • The Media Analysis API has been added. This API allows you to perform multiple Media Analytics API functions on the same media file.

  • The Image Recognition API no longer returns matches with a confidence less than 60%.

  • Performance of the Search APIs has been improved.

  • NORMALIZED_TO, CC and BCC fields in Explorer, Standard, Jumbo and Custom Fields flavor Text Indexes are now of type Parametric and Index.

  • The Parametric fields limit for the Custom Fields flavor has been increased from 10 to 30.

  • The root_parent_reference field has been added to all Text Index flavors with a field type of Reference.

  • A reference_field parameter has been added to the Get Content API to allow you to choose which Reference field to use.

  • The Extract Text API now populates the following new fields: root_parent_reference, doc_title, root_parent_title, keyview_version.

Haven OnDemand Developer Website Improvements

  • The API Key Management page has been improved. You can now give you API keys a name and description as well as refresh or delete them.

  • The list of news datasets used in the English and Italian News Text Indexes have been updated.

Combinations Improvements

  • Combinations now have versions associated with them.

  • A new Lowercase String utility is available for Combinations.

  • Combinations can now be imported and exported.

Resolved Issues

  • The List Resources API would not filter by flavor correctly.

  • The Document Categorization API would allow you to use Text Indexes which were not of the Categorization flavour.

  • The Restore Text Index API would return an unhelpful response if an invalid date was entered.

  • The Document Categorization API would return error 1000 (Unknown) if the index parameter was omitted.

  • Attempting to reset your authentication could cause an error.

  • The SharePoint Cloud Connector could fail to authenticate when using federated authentication.

  • The Usage and Quotas and API Usage page could be inaccessible if your API Units quota had been exceeded.

  • Query Manipulation Rules would not work correctly when phrases were used in synonym_add and synonym_remove.

  • The website would require you to re-authenticate after 30 minutes even if you were still active.

  • The API usage graph on the API Usage page would only show the first 1000 API calls.

  • The Combinations canvas would allow you to drag components outside of the visible area.

  • Combinations could be created without their parameter names being defined.

  • When editing a Combination the loading spinner could appear outside of the visible area.

4 January 2017

This section describes the enhancements and resolved issues in the January 2017 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Update Connector API has a new remove_credentials parameter. This parameter allows you to remove optional credentials from a connector configuration.

  • The Retrieve Config API has a new include_defaults parameter. This parameter allows you to return the default values in the configuration. This parameter is available only when you set output to config_attributes.

  • The configuration for the Web_cloud flavor connector has been updated, with the following changes:

    • The default value for url_cant_have_regex is now .*\.css$|.*.css\?.*$|.*\.js$|.*\.js,.*$|.*.js\?.* to exclude CSS and javascript by default. This change gives cleaner index results.

    • The spider_url_cant_have_regex and spider_url_must_have_regex configuration options were added, to allow you to control the pages that the connector crawls for links.

    • The content_type_cant_have_regex and content_type_must_have_regex configuration options were added, to allow you to allow or disallow particular content-type headers for the pages that the connector indexes.

    • The login_selector and password_selector configuration options were added. These parameter replace login_field_value and password_field_value. These new values allow you to specify a full CSS connector to select the user name and password elements to populate on a Web page.

    • The login_field_value and password_field_value parameters have been deprecated. This configuration allowed only <input> elements with a name attribute to be populated with user name and password values. Haven OnDemand recommends that you use the more flexible login_selector and password_selector options. These parameters are still available for existing implementations, but it might be incompatible with new functionality. The parameters might be deleted in future.

Haven OnDemand Developer Website Improvements

  • The Developer Website now includes a user message inbox to display notification messages about Haven OnDemand and your account.

  • The Haven OnDemand Pricing pages have been updated with additional information about rate limits for different plans.

  • The sign up processs has been improved.

Combinations Improvements

  • The Combinations editor has been improved so that the canvas UI does not allow you to create connections that are not valid.

  • A new AppendString utility is available for Combinations.

Resolved Issues

  • Quota errors were not returning additional details.

  • The Auto Complete API returned a 5000 error when the input text was whitespace.

  • The Combination icon was not displayed correctly on the Publish tab.

  • If a text index name contained a newline character, the index could not be deleted from the Haven OnDemand account management pages.

  • Attempting to create a text index with an invalid flavor returned an error 1000 (Unknown).

  • The platform APIs could sometimes return an error 1000 (Unknown) when an internal job failed.

  • Updating a developer profile from the Haven OnDemand management page could fail.

  • Creating a Combination with certain combinations of nested inlines could fail with an invalid combination error in the UI.

  • A Combination that contained a reference to a text index could fail with an error 5000 (backend request failed) if the index name specified was not valid.

  • The sign up verification process could reset if the verification email had to be resent.

  • The user Welcome page was not shown after first sign in.

  • The link for resending a verification email was hidden on the login screen.

  • The Connector History Try page allowed you to select multiple text indexes when only one was allowed.

  • The Update Connector API could return an error message if updating to the same values. It now returns success with a message.

  • The Cancel Schedule API could return an error message if a schedule did not exist or had already been cancelled.

22 November 2016

This section describes the enhancements and resolved issues in the November 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • Several new Connectors have been added to Haven OnDemand, for text indexing from different repositories:

    • Exchange Cloud Connector. Indexes content from your Exchange Online user mailboxes.

    • IMAP Cloud Connector. Indexes content from your IMAP email mailboxes.

    • Jive Cloud Connector. Indexes content from your Jive community.

    • Twitter Cloud Connector. Indexes content from Twitter.

    • SharePoint Cloud Connector. Indexes content from a SharePoint repository.

    • Web Onsite Connector. Indexes content from a Web address.

    For more information, see Connectors.

  • The Speech Recognition API now supports several new language options:

    • Australian English, telephony

    • Canadian English, broadband

    • Singapore English, broadband

    • South African English, telephony

    • Canadian French, broadband and telephony

    • Japanese, telephony

    • Dutch, telephony

    • Polish, broadband and telephony

    • Romanian, broadband and telephony

    In addition, the telephony language packs for Canadian English and American English have been improved.

  • The performance of the Speech Recognition API has been improved.

  • The Speech Recognition API now costs one API unit for every two seconds of audio processed.

  • The Train Prediction v2 API has a new selection_strategy parameter, which allows you to select the measurement criterion that defines the best algorithm for your prediction models.

  • The Update Connector API now returns an appropriate error message when attempting to use the invalidate_previous_state parameter for an onsite flavor connector.

Combinations Improvements

  • You can now update the display name and description on a published Combination.

  • The Combination creation wizard and canvas now load more quickly when switching tabs and saving, and is more responsive when moving items around the canvas.

  • Error and warning messages have been improved.

  • The UI now highlights required fields more clearly when they are empty when you attempt to test or publish a Combination.

  • Combinations now return details about a failing element when the Combination fails.

  • The Combination creation wizard now requires Combination names to be all in lower case when you create the Combination, rather than converting upper case characters to lower case.

Changes to Existing Functionality

  • The Update Connector API now replaces the existing configuration with the new one, rather than attempting to merge existing values when not specified.

Resolved Issues

  • The Combinations editor allowed users to insert the current combination.

  • On the Combinations canvas, it was not possible to add new inputs directly to an inline.

  • On the Combinations canvas, connections within nested inlines could be lost, resulting in an input error.

  • The Combinations Remove Key utility could cause an array to nest inside itself.

  • In Combinations, when a constant was added to a Boolean input, the input type was not set correctly.

  • The Hello World Public Combination was not treated as a valid inline Combination in the Traverse utility.

  • The Combination description could contain the string 'null' if no description was added.

  • Several minor display issues were fixed in the Combinations UI:

    • Default input values could disappear off the edge of the canvas.

    • The editor progress bar overlapped the buttons in narrow windows.

    • The 'add component' buttons were displayed on the read-only View Inside tab for published combinations.

  • The Accounts text index page listed resources that were not text indexes.

  • The JSON output for warnings in API results converted the reason and reference properties into a string. The format now matches the documentation at Warning Codes.

  • The Get Parametric Ranges API did not return results from private text indexes.

  • The Recommend API ommitted records from the response object when there was no recommendation.

  • Create Text Index API requests could take a very long time in certain circumstances, which could result in a timeout for synchronous requests.

  • The Stop Connector API would return an error if the connector was already stopped.

  • The View Document API would sometimes fail with error 5000 (Backend request failed).

  • The Start Connector API would sometimes fail with error 5000 (Backend request failed).


19 October 2016

This section describes the enhancements and resolved issues in the October 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Combinations feature is now available for all Haven OnDemand users. Combinations allow you to create workflows that include multiple Haven OnDemand APIs, without the need for you to write any code. You can then call one or more Combinations from your applications. Combinations include:

    • the Combinations Marketplace, which provides the list of available Combinations.

    • the Canvas user interface, which allows you to create, adjust, and publish your Combinations.

    • many utilities to perform simple functions within a combination, such as string extraction, and conditional logic.

    • several public Combinations, which Haven OnDemand has created for common uses. You can call the public Combinations directly, or use them in your own Combinations.

Resolved Issues

  • The Tokenize Text API returned a 5000 error when querying multiple indexes.

  • Some minor errors on the pricing pages have been corrected.


26 September 2016

This section describes the enhancements and resolved issues in the September 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Get Parametric Ranges API has been added. This API is similar to Get Parametric Values, and allows you to return search filters for a field that contains a continuous range of numeric values (rather than a discrete set of values). You can define a set of ranges, and the API returns the number of results where the specified field has a value within each range.
  • A new version of the Analyze Sentiment API is available. This API has an improved response format, and the aggregate score now returns new values mixed, slightly positive, and slightly negative for scores close to zero. The new mode parameter also allows you to prioritize precision or recall in your responses. The API documentation displays the latest version by default. To call this API, use v2 as the API version in the call. For example, http://api.havenondemand.com/1/api/sync/analyzesentiment/v2.
  • The Get Parametric Values v2 API has a new nest_field_results parameter. When you specify multiple values for field_names, you can set nest_field_results to true to find sets of results that occur together in documents.
  • Error reporting has been improved for the Image Analysis APIs.
  • The List Resources API can now return information about prediction model resources that were created with the v2 Train Prediction API.
  • The Combinations UI and stability have been improved. This functionality is in early access, but you can sign up to see what's coming at Combinations.

Platform Improvements

  • Platform performance has been improved.

Resolved Issues

  • The Recommend API could fail for invalid JSON files, rather than returning an appropriate error.
  • The Add to Text Index API Try It page did not correctly reset content in the form when the input type was changed.
  • The Query Text Index API could return only a single result when check_spelling was set to autocorrect.

24 August 2016

This section describes the enhancements and resolved issues in the August 2016 update of Haven OnDemand.

API Enhancements

  • In the Query Text Index API, error messages have been improved. Error 4024, "Field_text contains operations on unoptimized fields", now also covers the case of unknown fields. If there is more than one error in the query parameters, the one returned corresponds to the first in the order of the parameters in the index.
  • The JSON response schema for Get Parametric Values API V2 has been improved to say which fields are required, and to be more descriptive.
  • The Find Related Concepts API has been improved to filter out less relevant results.

Haven OnDemand Developer Website Improvements

  • Several APIs (Image Recognition, Restore Text Index, Retrieve Index Fields, Delete Query Profile, Retrieve Query Profile, Update Query Profile and Find Related Concepts) have a new quickstart section on the overview page, to help you get started.
  • New documentation has been added, including a Primer on How to Use the APIs, and a How-to for setting up a Dropbox Connector.
  • The File System Connector documentation has been updated to reflect a new installer.
  • The Web Cloud Connector How-to has been revised to include new sections on credentials, regular expression formulation and page clipping.

Platform Improvements

  • Platform stability has been improved.
  • Connector performance has been improved.

Resolved Issues

  • The Extract Entities and Analyze Sentiment APIs could fail with a 5000 error 'Backend request failed' when submitted in jobs with multiple files.
  • The Get Parametric Values API could fail with a 5000 error 'Backend request failed' when an invalid Boolean parameter was passed to the API.
  • The Retrieve Index Fields V2 API did not return date fields correctly.
  • The View Documents API could return an error on highlighting when used with a NOT operator.
  • Some API calls referencing resources, typically when the resources had long names, could fail with an error.

25 July 2016

This section describes the enhancements and resolved issues in the July 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • A new version of the Get Parametric Values API is available. This API has a new response format, and date values now return in ISO-8601 format. The field_name parameter has been renamed to field_names and now accepts multiple values. You can also specify the new date_period parameter to group date values into sets of different sizes. The API documentation displays the latest version by default. To call these APIs, use v2 as the API version in the call. For example, https://api.havenondemand.com/1/api/sync/getparametricvalues/v2.
  • New versions of the Train Prediction, Predict, and Recommend APIs are available. These APIs have a simplified input data format, and updated response format. The new versions of the Train Prediction and Predict APIs also support regression type algorithms, as well as the existing classification prediction type. Note that the Recommend v2 API does not yet support recommendation for regression type prediction models. The API documentation displays the latest version by default. To call these APIs, use v2 as the API version in the call. For example, https://api.havenondemand.com/1/api/sync/trainpredictor/v2.
  • You can use the List Resources API to list prediction models that you create using the v2 Train Prediction API.
  • The Get prediction Model Details API has been added. You can use this API to get the prediction model performance measures, and the structure of the data set that you used to train the model. You can retrieve details only for prediction models that you trained by using the v2 Train Prediction API.
  • The Delete Prediction Model API has been added to allow you to delete an existing prediction model that you no longer require. You can delete only prediction models that you create by using the v2 Train Prediction API.
  • The Tokenize Text API has a new ignore_operators parameter to allow you to tokenize the input text as plain text, even if it contains query operators.
  • The Tokenize Text API now returns the length and case response parameters for all terms.
  • The Recognize Images API corporate logos dataset has been updated. For a full list of available corporate logos, see Corporate Logo Training Set.
  • Error reporting has been improved for various APIs, particularly for cases where the input parameters or files did not contain valid content.

Haven OnDemand Developer Website Improvements

  • New documentation pages are available, about using GET and POST methods, and API Versioning in Haven OnDemand.
  • Several APIs (Find Similar, Auto Complete, Connector Status, and Cancel Connector Schedule) have a new quickstart section for the overview page, to help you get started.

Changes to Existing Functionality

  • For the Train Prediction, Predict, and Recommend v1 APIs, prediction services are now referred to as prediction models, in line with the new v2 APIs. The service_name parameter has been renamed to model_name. The service_name parameter is still available, for backwards compatibility. However, Haven OnDemand recommends that you update your application to use the latest version of the API.

Resolved Issues

  • The Get Content API did not respect the value specified in the end_tag parameter for highlighting.
  • The Get Content API did not respect the start_tag if the value was not valid XML and there was no end_tag specified.
  • The Sentiment Analysis API could fail when multiple files were selected in the Try It page.
  • The Add to Text Index API could fail a batch of documents if a single document had an import error.
  • The Add to Text Index API could fail to index the contents of a container file if one of the files was not valid.
  • The Auto Complete API returned a 5000 error when the input text contained 4-byte UTF-8 characters.
  • The View Document API could render charts in Microsoft Excel charts incorrectly.
  • The API Try It page option to render carriage returns instead of \n was not visible.

11 July 2016

This section describes the enhancements and resolved issues in the July 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Sentiment Analysis API has been improved for the English language. In particular, topic extraction has been improved.
  • Error reporting has been improved for text indexing APIs when running an API against multiple text indexes.
  • The Filesystem Onsite flavor connector now supports options for Mapped Security. To take advantage of this feature, Contact Us.

Resolved Issues

  • In some Web browsers, input parameters on the API Try It pages could attempt to fill in passwords.
  • The output from the View Document API could be different depending on whether the API was called with HTTP or HTTPS.
  • Some APIs returns a 5000 error (backend request failed) for an invalid file type. The APIs now return an appropriate error message.
  • The Extract Text API response schema incorrectly listed content-type as a response property, when the API returns content_type.

28 June 2016

This section describes the enhancements and resolved issues in the June 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Discovery platform API now retrieves API definitions.
  • The Extract Entities API now returns a more appropriate error response when the text parameter is empty.
  • Error responses were improved for the View Document API.
  • The performance for the Unstructured Text Indexing APIs has been improved.

Haven OnDemand Developer Website Improvements

  • A new client libraries page has been added to the developer website.

Resolved Issues

  • Some API Try examples returned an insufficient privileges error when the user was not logged in.

06 June 2016

This section describes the enhancements and resolved issues in the June 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The License Plate Recognition API has been added. This API recognizes license plates in a video file, and provides some information about the license plate.
  • The Detect Scene Changes API has been added. This API detects frames in video in which the scene changes significantly from the previous frame, and returns these images.
  • The Text Statistics API has been added. This API analyzes text and returns information about the content, such as the number of sentences and terms.
  • The Image Recognition API now returns a confidence value for the recognized image.
  • The dictionary for the Auto Complete API has been improved.
  • Error reporting has been improved for unsupported file formats and syntax errors.
  • Query performance has been improved for queries that include a large number of text indexes, or large text indexes.
  • The Web Connector has been updated to download more relevant content, and its performance has been improved.
  • The Spanish language pack has been improved for the Speech Recognition API, for improved recognition accuracy and to correct some pronunciation errors.
  • The Retrieve Index Fields API performance has been improved for text indexes with lots of fields.
  • The Sentiment Analysis API has been improved for the English language option. In particular, the API now includes sentiment recognition for emoji, and improved scoring for uppercase terms.

Haven OnDemand Developer Website Improvements

Changes to Existing Functionality

  • In the Entity Extraction API, the names of a number of entity types have been changed for consistency. The previous name for these entity types is still available, for backwards compatibility. However, these entity types will be deprecated in the future, and Haven OnDemand recommends that you update your applications to use the latest version:
    • organizations is now called organizations_eng
    • languages is now called languages_eng
    • professions is now called professions_eng
    • universities is now called universities_eng
    • profanities is now called compliance_eng
    • films is now called films_eng
    • teams is now called teams_eng
    • holidays is now called holidays_eng
    • medical_conditions is now called medical_conditions_eng

Resolved Issues

  • The documented limit for the sample_size parameter in the Find Related Concepts API was incorrect.
  • Exceeding the allowed range for the sample_size parameter in the Find Related Concepts API did not return an error.
  • The API try pages could take a long time to show large JSON responses.
  • The Update Connector API did not work correctly with onsite connectors.

05 May 2016

This section describes the enhancements and resolved issues in the May 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Map Coordinates API has been added. This API maps a latitude and longitude position that you provide to a country, timezone, or zipcode (in the USA).
  • The List Resources API now returns the display name for resources.
  • The precision of the Face Detection and OCR Document APIs has been improved.
  • The Face Detection API has a new face_orientation parameter to specify information about the faces in the image that you provide.
  • The OCR Document API has a new languages parameter to specify the language of the text, to improve the accuracy of the results. For a list of supported languages, see Supported OCR Languages.
  • Support has been added for an on-premise user connector for Enterprise user imports.

Haven OnDemand Developer Website Improvements

Resolved Issues

  • The Extract Entities Try page did not escape HTML.
  • The corporate logos public dataset for the Recognize Images API did not include the Hewlett Packard Enterprise Logo.
  • When viewed on an iPad, the entity_type list box for the Extract Entities API could not be closed.
  • If the login page was open for more than 30 minutes, an attempt to log in could fail.
  • The Face Detection API results were only available to logged-in users.
  • The Connector Status API could fail for old connectors.
  • An asynchronous request to the Add To Text Index API could return a timeout after five minutes.
  • The Recognize Speech API could miss the first few seconds of the audio.

29 March 2016

This section describes the enhancements and resolved issues in the March 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Speech Recognition API now supports the following additional language packs:
    • Modern Standard Arabic
    • Australian English
    • Canadian English
    • Latin American Spanish Telephony
    • Farsi (Persian)
    • Japanese
    • Brazilian Portuguese
    • Russian
  • The Create Text Index Try It page now displays the unit cost for each index flavor.
  • Haven OnDemand now supports SAML for corporate login.

Haven OnDemand Developer Website Improvements

  • A new Getting Started page has been added for new users.
  • Various minor improvements and usability enhancements have been made to the developer Website.

Resolved Issues

  • The Query Text Index API could sometimes handle square brackets incorrectly.
  • The Delete Text Index API Try It page could list a recently deleted text index.
  • The Query Text Index API sometimes failed when using check_spelling with field_text.

03 March 2016

This section describes the enhancements and resolved issues in the March 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • New HP Labs APIs are now available. These APIs provide new data analytics functionality:
    • Anomaly Detection. The Anomaly Detection API allows you to detect anomalous records in a dataset.
    • Trend Analysis. The Trend Analysis API identifies the most important changes between two similar datasets (for example, sales data for two different time periods).
  • A new text index flavor, Jumbo, is available. This text index flavor has a similar configuration to the Standard flavor, but it provides 40GB of document storage for larger indexes.
  • The Standard flavor text index size quota has been increased to 10GB document storage.
  • The Speech Recognition API now has language options for Dutch (NLNL) and Latin American Spanish (ESLA).
  • New versions of the onsite Connector installer are now available.
  • Haven OnDemand now supports mapped security for Hybrid deployments.
  • API response times have been improved.

Haven OnDemand Developer Website Improvements

  • Various improvements have been made to the developer Website user experience.

Resolved Issues

  • The Create Query Profile API query_manipulation_index parameter did not accept text index names that included capital letters.
  • The Search APIs (Query Text Index, Find Similar, and Get Content), could return an empty array for the content of a document when the item in the index was not text.
  • In some cases, it was not possible to log in to the Haven OnDemand website using Internet Explorer.

18 January 2016

This section describes the enchancements and resolved issues in the January 2016 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Recognize Speech API now returns a confidence value, and the duration (in milliseconds) for each extracted word in the transcript.
  • The Recognize Speech API now supports several telephony language options, to allow you to process speech from telephone calls.
  • A new Transport public text index is available. This text index includes information about various transport hubs, such as train stations and airports.
  • Error responses for the Get Parametric Values API have been improved.
  • Error reporting for text indexing has been improved.

Haven OnDemand Developer Website Improvements

  • The login process has been improved.

Resolved Issues

  • When using query profiles, the Get Parametric Values API returned results in inconsistent cases, which could lead to duplicate results.
  • For APIs that accepted an array of inputs, the file input was not listed correctly, even though it was passed to the API.
  • The Services Text Indexes section the Accounts page sometimes failed to display results for the Query Text Index interface.
  • An error could occur when deleting a text index that spanned multiple index components.
  • Cancelling the login process could lead to a site error.
  • Attempting to verify a signup that has already been verified could return a HTTP 500 error.

14 December 2015

This section describes the enchancements and resolved issues in the December 2015 update of Haven OnDemand.

API Enhancements

  • Language processing for the Recognize Speech API has been updated to improve the results.
  • Various improvements have been made to API and platform documentation.

30 November 2015

This section describes the enchancements and resolved issues in the November 2015 update of Haven OnDemand.

New Functionality and Enhancements

  • New versions of the Entity Extraction and Retrieve Index Files APIs are available. The API documentation displays the latest version by default. To call these APIs, use v2 as the API version in the call. For example, https://api.havenondemand.com/1/api/sync/extractentities/v2.
  • The Entity Extraction API (version 1 and 2) has been updated to return new additional information fields for the medical_conditions, universities, teams, films, languages, and holidays entity types.
  • The Entity Extraction API version 1 response has been updated to display an offset, which provides the location of the match in the original text.
  • Various improvements have been made to the Prediction APIs.

    • The Prediction APIs now provide a confidence calculation for predictions.
    • The validation for prediction data has been improved.
    • Error reporting has been improved.
    • The Predict API now supports missing values.
    • The Prediction APIs now validate empty and unique header names.
    • Stability and availability has been improved.

Haven OnDemand Developer Website Improvements

  • The API Try pages now provide curl examples for both GET and POST requests.
  • The API documentation now displays multiple versions of APIs. When you view older versions of the API, the API pages display a note that a newer version is available.
  • The login and sign up process for Haven OnDemand has been improved.
  • Passwords provided to APIs in calls are now masked in the account Request Log.

Resolved Issues

  • On signup, the Captcha image did not reset after a signup error.
  • When two Haven OnDemand user accounts used the same social media for authentication, resetting the provider on one account could prevent access to the other.
  • When a text index name contained an ampersand (&), errors could occur when modifying the text index from the Account Management page.
  • The Connector History API connectors parameter drop down did not work correctly on the API Try page.
  • The Query Text Index API Try page text input box could be disabled after a call that resulted in an error.
  • Input boxes for the API Try pages were sometimes not cleared for subsequent calls.
  • The curl example on the API Try pages sometimes contained an undefined file name.
  • The Prediction APIs could return incorrect line and column numbers in error messages.
  • The Prediction APIs could return an incorrect error message when they received invalid JSON.
  • The Train Prediction API could fail when it received valid JSON that did not match the schema.

12 November 2015

This section describes the enhancements and resolved issues in the November 2015 update of Haven OnDemand.

New Functionality and Enhancements

  • The new Autocomplete API has been added. This API allows you to expand a specified prefix to a list of matching possible terms, for example to allow you to autocomplete search criteria.
  • The Query Text Index API now supports a spellcheck option.
  • The Update Query Profile API now allows for partial updates of a query profile.
  • The Web Connector now supports automatic web clipping.
  • The Recognise Speech API now returns up to 30 minutes of text. Previously, it was limited to five minutes.
  • The job result API now returns a more meaningful error message when it times out. It returns an error code of 7010, with the message "API request timed out. For long running requests, please use the async API."
  • JSON responses from the APIs are now properly formatted to be more human-readable.

Haven OnDemand Developer Website Improvements

  • The login process has been improved.
  • Login is now preserved when you click on a highlighted blog article.
  • Various other minor usability improvements have been made.

Resolved Issues

  • The SharePoint Connector could fail to crawl non Office-365 SharePoint sites.
  • Some users were unable to authenticate using HP Passport.
  • Asynchronous jobs failed if they contained four-byte UTF-8 characters.

September 2015

This section describes the enhancements and resolved issues in the September 2015 update of Haven OnDemand.

New API Functionality and Enhancements

  • A preview of the Graph Analytics APIs is now available. These APIs provide a number of analysis functions against a knowledge graph derived from the links between English Wikipedia pages. The APIs include:

    • Get Common Neighbors
    • Get Neighbors
    • Get Nodes
    • Get Shortest Path
    • Get Subgraph
    • Suggest Links
    • Summarize Graph
  • Two new connector flavors are available:

    • SharePoint Remote Onsite. Allows you to index content from a SharePoint server.
    • Dropbox Cloud. Allows you to index content from a Dropbox application.
  • The following additional fields are now Parametric by default in the Standard and Explorer text index flavors:

    • "CATEGORY"
    • "SOURCE_CONNECTOR"
    • "SOURCE_REPOSITORY"
    • "NORMALIZED_AUTHOR"
    • "NORMALIZED_FROM"
    • "ENRICHED_COMPANY"
    • "ENRICHED_PERSON"
    • "ENRICHED_PLACE"
    • "AUTHOR"
    • "FROM"
    • "ROOTPARENT_FROM"
    • "CONTENT_TYPE"

    In addition, a new default numeric field has been added, FILE_SIZE.

  • Significant improvements have been made to the Query Profile APIs, including updated documentation.

    In addition, the Retrieve Query Profile API has been added.

  • Haven OnDemand now supports GZIP compression of the body, if the Content-Encoding header is set to gzip.

  • Haven OnDemand now supports GZIP responses if the appropriate Accept-Encoding header is sent.

  • The Get Content API now has a summary parameter.

  • The List Resources API now displays query profiles.

  • The Entity Extraction API now returns additional fields for company entities.

  • The Sentiment Analysis and Entity Extraction APIs now support passing in multiple documents.

Changes to Existing Functionality

  • Changes have been made to the output of the Extract Text API to provide improved consistency and increased clarity.

  • The following fields have been renamed:

    • content-type to content_type.
    • original_size to file_size.
  • The following fields are no longer returned:

    • tempextractionfilename
    • keyview_type
    • keyview_class
  • Performance Improvements have been made to Speech Recognition API.

  • The Find Related Concepts API default sample_size parameter has been changed from 50 to 250.

  • The number of Parametric and Numeric fields that can be created as part of the custom_fields flavor, has been increased to 10 in each case.

Resolved Issues

  • The Retrieve Index Fields API could fail when used against certain indexes.
  • The Entity Extraction API original_length now corresponds to the original_text instead of normalized_text
  • The following issues have been resolved with the Predictive Analytics APIs:

    • Stability has been improved for the Train and Predict APIs.
    • Error messages were improved to contain more useful information.
    • The performance of the Train API has been improved.
  • The Speech Recognition API returning the wrong result or only one sentence for minutes of text.
  • Haven OnDemand automatically closed connections after a predefined period, regardless of whether the connection was in use. Now connections stay open when users are sending large files or to return large results.

Other Changes

As part of this release we are making significant changes to the underlying platform to support new functionality. Be sure to check out our next release for more details.


March 2015

This section describes the enhancements and resolved issues in the March 2015 update of Haven OnDemand.

New API Functionality and Enhancements

  • New Query Manipulation APIs have been added. These APIs allow developers to influence the content and results of a query against a text index, by defining a query profile. Examples of Query Manipulations include promotions, where additional results are injected into the result set, and synonyms, where given search terms can be expanded to include synonymous terms. For details see the Create Query Profile, Update Query Profile, and Delete Query Profile APIs.

  • The List User Roles API has been added to display the roles associated with a given user.

  • New Policy Management APIs have been added. These APIs provide an abstracted layer on top of entity extraction, categorization and related functionality to allow you to define policies to manage information. You can create classifications, to group documents and other objects into collections. The Policies then perform actions, such as indexing, on the documents associated with a collection.

  • The Recommend API has been added. This API uses your existing prediction models to recommend changes that you must make in a set of input values to achieve a desired prediction result.

  • The wording and formatting of some error messages has changed for increased consistency.

  • All Haven OnDemand endpoints now support the OPTIONS header.

  • The Get Parametric Values API now requires you to specify field names. You cannot use an asterisk (*) as a wildcard. This change prevents an inadvertent return of a large dataset.

  • The Get Parametric Values API now returns values for a date field grouped by hour.

  • The Find Related Concepts API now support multiple indexes.

  • The Extract Entities API now normalizes additional date formats.

  • The performance of the unstructured text indexing APIs have been improved.

  • The Stop Connector API has been added to allow you to manually stop a running connector.

  • The Cancel Connector Schedule API has been added to allow you to cancel an existing connector with a configured schedule.

New Haven OnDemand Developer Website Functionality

  • A number of minor enhancements to website functionality have been made, including additional documentation.

Resolved Issues

  • The Retrieve Index Fields API returned a DOCUMENT/ prefix to fieldnames. The returned field names are now consistent with other APIs.

January 2015

This section describes the enhancements and resolved issues in the January 2015 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Recognize Speech API has been renamed to the Speech Recognition API.

    The Speech Recognition API (recognizespeech) now supports a number of additional languages:

    • German
    • French
    • Italian
    • European Spanish
    • Mandarin

    Haven OnDemand recommends that the you use the asynchronous API for Speech Recognition because of its processing times.

  • The Prediction APIs have been added. These APIs allow you to train a prediction model and get the most optimized model for your data, without requiring ML or algorithm knowledge. After you train a model, you can run a prediction on new unclassified files and export the results.

  • The Authenticate API now returns a token. In future releases, this token will be supported in place of API keys to allow a combined session token for an application and an individual user.

  • New Platform APIs have been internally designed. These general APIs include Job Result, Job Status and API Discovery. They will be documented through the API Explorer in future releases.

  • The Start Connector API now accepts the ignore_previous state parameter to allow you to reindex all content, regardless of previous connector runs.

  • The Text Extraction API now supports a extract_xmlattributes parameter, to allow you to extract XML attributes from the file.

  • Connector schedules for the Create Connector API now support additional date formats.

  • The Categorization text index flavor now accepts the match_punctuation parameter.

  • The Query Text Index and Get Parametric Values APIs now support querying and returning results from multiple text indexes in a single call.

  • The Get Parametric Values API can be used on Numeric Date fields, with values aggregated on an hourly basis.

New Haven OnDemand Developer Website Functionality

  • A number of improvements have been made to options for password and account resetting.

  • The Try pages now provide improved context for resource drop-downs.

  • The Try pages now support calling both the synchronous and asynchronous versions of each API.

  • Various refinements have been made to the documentation and content.

Resolved Issues

  • The List Stores API could occasionally return inconsistent results

  • Using an asterisk (*) as the value in the field_name parameter of the Get Parametric Values API could cause performance issues. This value is no longer supported.

  • Certain text index names could cause issues with index scaling

  • A HTTP OPTION request could result in a stack trace being returned to the caller.

  • The Date field was not correctly configured for the standard and explorer text index flavors. It is now supported as a date field.



November 2014

This section describes the enhancements and resolved issues in the November 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • An initial preview of some of our User Management APIs is now available, allowing you to create, read, and authenticate users.

  • The Create Connector API now supports a Filesystem_onsite flavor. This flavor allows you to download a connector to install locally for indexing local filesystems.

  • The Web connector flavor has been renamed to Web_cloud to enable a future release of a web_onsite flavor for crawling of non-public websites.

  • The List Resources API has been added. This supersedes the List Indexes API. It provides a more consistent mechanism for developers to retrieve their assets such as text indexes, connectors and user stores, regardless of type.

  • The following new Entity Extraction types have been added:

    • Profanity. Identifies language that would typically be considered inappropriate in general communication.

  • Other API enhancements and changes:

    • The limit for the absolute_max_results parameter for the Query Text Index and find Similar APIs has been increased from 100 to 2,500. For the Get Parametric Values API, the limit for this parameter has been increased to 5,000.

New Haven OnDemand Developer Website Functionality

  • Users can now use HP Passport accounts to authenticate with Haven OnDemand.

  • A number of pages have been redesigned to improve navigation.

Resolved Issues

  • In certain situations, the error code 1000 (Unknown Error) was returned in place of 4030.

  • Createconnector API calls would fail to create the connector where the destination text index was created by the same user, but under a different API key.

  • Addtotextindex API calls failed if a reference was not provided.

  • The Web Connector stopped fetching content when it encountered invalid headers.

  • Including the Apikey parameter twice resulted in a quota error message.



October 2014

This section describes the enhancements and resolved issues in the October 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Extract Document Concepts API has been added. The Extract Document Concepts API allows strong concepts to be extracted from a document based on statistical methods. This allows concepts to be identified even in domain specific documents as it does not rely a predefined list.

  • The Recognize Speech API has been added. The Recognize Speech API allows developers to provide an audio file and have a transcript returned. The Recognize Speech API supports British and American English.

  • The Retrieve Index Fields API has been added, this allows users to query the fields and their corresponding types that are present in a text index.

  • The following new public text index datasets have been added:

    • Arxiv provides an index of scientific papers released through arxiv.org.

    • News_ger provides an index of News abstracts from a number of German language news sources.

  • The following new entity extraction types have been added:

    • The Film entity extraction type allows users to identify and extract the names of over 320,000 films from documents.

    • The File_hash entity extraction type extracts 32 and 40 character hexadecimals.

  • Additional Web Connector Parameters

    • Haven OnDemand Web Connector supports a number of additional configuration parameters to provide increased spidering and indexing control, including depth, url_must_have_regex; max_page_size and min_page_size.

  • The Entity Extraction API now support the unique_entities parameter that leads to entities referenced multiple times within a document being return just once.

Resolved Issues

  • Combining the PII Entity type with entities types that are a subset of the PII Entity Type could cause a 5000 error.

  • 4018 error could be return when passing a URL parameter that required cookies or a common user-agent.



August 2014

This section describes the enhancements and resolved issues in the August 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • Additional Entity Types have been added to the Entity Extraction API.

    • Universities. Matches over 2000 institutes of higher education.

  • The performance of the Add to Text Index API has been improved. When performing multiple Add to Text Index calls, additional parts of the process are now run in parallel to improve throughput.

  • The following new public text index datasets have been added:

    • World_factbook. Includes information from the 2014 CIA World Factbook.

    • Patents. Includes US Patent fillings from 2002 to 2014

  • The API Discovery API has been added (/1/discovery/api). This API returns a JSON-based list of all the APIs that you have access to in Haven OnDemand.

  • The Entity Extraction API datasets have been improved. The people dataset has been updated include additional individuals.

  • The corporate_logos dataset has been improved.

  • A number of APIs have been added for the provisioning and management of cloud connectors:

    • Create Connector. Creates a connector configuration that then either runs on a schedule, or after a call to the Start Connector API.

    • Delete Connector. Deletes an existing connector configuration.

    • Update Connector. Allows you to make changes to an existing connector configuration.

    • Start Connector. Runs one of your connector configurations.

  • The Categorization flavor has been added for text indexes. The categorization flavor allows you to create indexes that you can use to profile and categorize documents.

  • The Categorize Document API has been added. This API allows you to compare a document to the categories defined in a text index with the Categorization flavor.

  • Other API enhancements and changes:

    • The Language Identification API now supports an addn_information parameter. If you enable this parameter, the API returns additional information on the number of native speakers as well as the related Wikipedia entry and an associated graphic.

    • The Entity Extraction API now supports multiple entity types as part of a single call.

    • The Get Parametric Values API now supports a max_results parameter to control the number of concepts returned.

    • The Query Text Index, Highlight Text, Find Similar, View, and Get Content APIs now support the end_tag parameter. You can use this parameter to explicitly specify an ending tag to use for highlighting, in conjunction with start_tag.

    • The max_results parameter in the Query Text Index API has been renamed to absolute_max_results. start and max_page_results parameters have been added. max_results will be supported as an alias until further notice.

New Haven OnDemand Developer Console Functionality

  • The Text Index Creation Wizard on the Accounts page has been updated.

  • You can now search the Haven OnDemand site.

Resolved Issues

  • The Create Text Index API could fail in certain scenarios.

  • For a short period after the creation of a text index, calls to related APIs such as Add to Text Index could return error code 8002.



July 2014

This section describes the enhancements and resolved issues in the July 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • Additional entity types have been added to the Entity Extraction API.

    • Professions_eng. Matches over 13,000 professions in English.

    • Organizations. Matches over 26,000 organizations covering charities, political parties, religious bodies and inter-governmental agencies.

    • Drugs_eng. Matches scientific and common names for over 2,000 drugs.

    • Language. Matches over 6000 languages.

  • The Entity Extraction datasets have been improved.

    • The Place Name entity type now extracts a number of geographical features such as Lakes and National Parks.

    • The Place Name entity type now includes place_continent and place_type as additional_metadata.

    • The Notable_People entity type now identifies over 100,000 individuals.

    • The Entity Extraction API now returns an image tag under additional_metadata for places and people where available.

  • The Face Detection API now now has improved recall.

  • The Add To Text Index supports multiple file, additional_metadata and reference_prefix sets.

  • The Web Cloud Connector has been added, on an Early Access program.

    Haven OnDemand Web Connector allows user to spider a Website to index the content into a Haven OnDemand text index. To be one of the first users, contact the Haven OnDemand team.

  • Performance has been improved for synchronous API calls. The total request time has been reduced by 150ms on average.

  • Other API enhancements and changes:

    • The Start_Tag parameter for the Highlight Text, Query Text Index, Get Content, Find Similar, and View Document APIs now has a default value of <span style=\"background-color: yellow\"> for improved usability.

    • The Reference_prefix parameter for Add to Text Index API now preserves case.

Haven OnDemand Developer Console Functionality

  • The Public Datasets page has been updated to include additional information.

  • The Text Index Creation Wizard has been updated to allow you to define custom parametric and index fields, where supported by the index flavor.

  • The documentation has been improved to include updated overviews for many APIs, including Quick Start examples.

Resolved Issues

  • Create Text Index API calls could fail in certain scenarios.

  • The error code 7004 was returned for two cases: "Too many fields in the request" and "Too many files attached to the request". The "Too many fields in the request" error now returns code 7001.

  • Using a numeric value for the reference in the Add to Text Index API could result in an error.



May 2014

This section describes the enhancements and resolved issues in the May 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • The Get Content API has been added. You can use this API to return the contents of a document from a text index based on a reference. You can also specify which fields return, and you can highlight based upon any valid Haven OnDemand Boolean expression.

  • The additional_metadata parameter has been added to the Extract Text and Add to Text Index APIs. You can use this parameter to pass a JSON object that contains a list of fields that you want to include in the document. The APIs include the additional values with the content extracted from the file.

  • The Recognize Barcodes API now supports QR Codes.

  • Unoptimized field_text is now rejected for the Query Text Index, Find Similar, Get Parametric Values and Find Related Concepts APIs.

  • The Standard Text Indexing Flavor how includes a MODIFIED_DATE date field and has a default Parametric field of CATEGORY.

  • The Explorer flavor has been added for text Indexes. The Explorer flavor matches the configuration of the Standard flavor, but with a reduced maximum number of document and total index size.

    flavor=explorer
  • New IP Address, Email Address, Address, Person Fullname, Person Name Component, and PII entity extraction types have been added.

    • The Address entity extraction types match full address blocks consisting of a street name and city.

      entity_Type=address_au
      entity_Type=address_ca
      entity_Type=address_de
      entity_Type=address_es
      entity_Type=address_fr
      entity_Type=address_gb
      entity_Type=address_it
      entity_Type=address_us
      entity_Type=address_zh
    • The Person Fullname entity extraction type matches First Name and Surname combinations.

      entity_type=person_fullname_eng
    • The PII (Personal Identifying Information) and PII Extended entity extraction types are supersets of a number of other Entity Extraction types that are commonly considered PII.

      entity_type=pii
      entity_type=pii_ext
  • The Entity Extraction datasets have been improved.

    • Person Name entity extraction has been refined to match known individuals in a wider set of cases.

    • Place Name entity extraction now recognises common abbreviations for places in its dataset, for example Ft. Collins for Fort Collins, and disambiguation has been improved.

    • The companies dataset has been extended to include over 45,000 companies including privately-held companies.

  • Other API enhancements and changes:

    • The database_match parameters have been renamed to indexes to improve consistency across the platform. database_match name will be supported as an alias until further notice.

    • The Get Parametric Values API now supports sorting by the number of matching documents for each value. sort=document_count

    • The Query Text Index API now supports a start parameter to allow paging of results.

    • The max_results parameter in the Find Related Concepts API has been renamed to sample_size for improved clarity. max_results will be supported as an alias until further notice.

    • The target barcode type can be passed to the Recognize Barcodes for increased performance. barcode_type=type

    • The OCR Document API now has a fourth mode, and the existing modes have been renamed to clarify the intended usage. The existing mode names will be supported until further notice.

  • The terms option has been added for the highlight parameter of the Query Text Index API, which highlights individual terms from the query.

Haven OnDemand Developer Console Functionality

  • The Haven OnDemand Showcase has been added to showcase user applications developed using Haven OnDemand. If you would like your application to feature, contact us.

  • The Documentation has been improved:

    • How To documents have been added to provide holistic overviews.

    • The overviews have been updated, including Quick Start examples

    • Many other minor improvements have been made

Other Improvements

  • Haven OnDemand now utilizes HPCS 13.5

Resolved Issues

  • The Text Tokenization API returned 0 for documents and occurrences.

  • The Image Recognition API sometimes failed to match images where a transparency existed.

  • The Delete Text Index API could fail in certain scenarios.

  • The Extract Text would return root as reference instead of filename.



April 2014

This section describes the enhancements and resolved issues in the April 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • The following updates have been made to Haven OnDemand dynamic indexing:

    • The Delete From Text API has been added. You can use this API to remove individual documents by their index reference.

    • The Index Status API has been added. You can use this API to retrieve the number of documents and the size of a given index, and the number of changes in the last 24 hours.

    • The Add to Text Index API now has support for preventing duplicate documents. When an existing document with the same reference exists, you can now choose whether to delete the existing document, or to index a duplicate.

    • Haven OnDemand now automatically creates backups and transaction logs and maintains them for all user indexes, in case of a failure.

    • The List Indexes API has been updated to return additional information.

  • The following updates have been made to Haven OnDemand Search APIs:

    • The Get Parametric Values API has been added. You can use this API to aggregate configured Parametric fields.

      For example, try a parametric search with index=wiki_eng and field_name=WIKIPEDIA_TYPE.

    • The Query Text Index API now supports creation of summaries during a query.

    • The Query Text Index API and Find Similar API now support summarization. See the summary parameter.

  • Face detection now provides age analysis based on four categories (child, youth, adult, and elderly).

  • The Entity Extraction API list of companies has been updated.

  • The Entity Extraction API list of places has been updated.

  • The Entity Extraction API now returns additional information, such as person_gender, person_profession, and so on, under an additional_information object in the JSON schema.

  • A number of API names have been changed for consistency and ease of use. For a full list, see the Appendix. Previous API names continue to work until further notice.

  • The following enhancements have been made to the public datasets:

    • New text indexes have been added for German, Chinese, and Italian Wikipedia.

    • Structured data has been added to the non-English Wikipedia datasets.

    • The Companies dataset has been updated.

    • The News dataset has been updated.

    • The News dataset is now available in Italian and French.

  • APIs, APIKeys, and text indexes now all have quotas. Quotas can be adjusted on a global, service level, and per-user basis.

Developer Console and Website Improvements

  • You can now use the Try It functionality without logging on to Haven OnDemand.

  • Haven OnDemand now supports Lithium Integration (SSO and Header).

  • The Service Status page has been added to display a summary of the service status of all APIs.

  • The unapproved user signup process now includes CAPTCHA.

  • The developer console account management section now includes text index creation and management pages.

  • The API navigator functionality has been improved. The API navigator now supports a list view with categories, to allow you to find APIs more efficiently.

  • The developer console now has cookie notifications to comply with EU privacy laws.

  • The following improvements have been made to the documentation:

    • Additional documentation has been added, based on developer feedback.

    • When viewing the request tab, inline links are now provided to related documentation.

    • New pages refer to the availability of onsite Connectors.

API Resolved Issues

  • S could be returned as part of query summary results

  • The word Harry was incorrectly specified as a stop word.

  • Image recognition examples specified invalid parameters, which caused the examples to fail.


Appendix

The following table lists the API name changes in this update of Haven OnDemand.

Previous API Name Current API Name
detectsentiment analyzesentiment
findfaces detectfaces
dynamicthesaurus findrelatedconcepts
explodecontainer expandcontainer
expandterm expandterms
extractentity extractentities
extracttext extracttext
findsimilar findsimilar
highlight highlighttext
detectlanguage identifylanguage
ocr ocrdocument
readbarcode recognizebarcode
recognizefaces recognizefaces
detectimage recognizeimages
storeobject storeobject
tokenize tokenizetext
view viewdocument
query querytextindex
addtotextindex addtotextindex
createtextindex createtextindex
deletetextindex deletetextindex
listtextindexes listindexes


March 2014

This section describes the enhancements and resolved issues in the March 2014 update of Haven OnDemand.

New API Functionality and Enhancements

  • Haven OnDemand now supports dynamic indexing. You can:

    • create a new index. Currently one flavor is available, which support for up to 250,000 documents.

    • delete an index.

    • index documents.

    • query documents and use dynamic thesaurus against your dynamic indexes

  • The Detect Language API can now match shorter language strings. For example:

    • This is English

    • Ich bin ein Berliner

    • What is this?

    • 最新的大约40万

  • The Detect Language API now returns ISO639-2B codes, as used as input by other APIs.

  • The Detect Image API database matching is now case insensitive, to be consistent with the Query API.

  • The Read Barcode API now returns additional country information for EAN-13 barcodes.

  • The Dynamic Thesaurus MaxResults parameter now has a default value of 50, instead of six, to provide better results.

  • Entity Extraction for people has been extended to include historical famous figures.

  • Entity Extraction for places has been modified to include US States, and to not require State names to match cities. In addition, place name disambiguation has been added.

  • Parameter names have been updated in several APIs to improve consistency. Review your code to ensure compatibility.

  • Core API code has been optimized to improve throughput and reduce latency.

  • Sentiment Analysis has been greatly improved for non-English languages.

API Resolved Issues

  • EAN-13 barcodes were always reported as ISBN, despite ISBN being only a subset. Bookland(ISBN) is now reported as part of the country information.

  • Stop lists were incorrectly applied to AgentBooleans.

  • The View API would accept files of any size, presenting a potential DoS risk. It now stops at 100MB.

Developer Console and Website Improvements

  • Various minor usability improvements have been made.

  • Non-logged in overviews now return the correct 404 error.

  • An issue with verification email security has been resolved.

  • Support has been added for experimental APIs.