Notas de publicación

Mejoras

• Language engines:
  • Updated language resources for the Hunspell spell check engine: British English, French, Indonesian, Dutch, Polish, Ukrainian, Mexican Spanish, Danish, German, Swiss German, Bulgarian, Czech, Chilean Spanish.
  • Slovenian language code was changed from sl_SL to sl_SI. This change affected the 2nd part of the locale ID with the country code. It’s SI (Slovenia). The old language code will still work for the existing clients. However, it’s recommended to make changes in the configuration and migrate to the correct language code.
  • Prioritized grammar type suggestions if the spelling engine gives “No suggestions”. In other cases, spelling suggestions will supersede.
  • [On-premise] Updated the structure of the folders/directories with resources such as language files and additional wordlists.
• User-level custom dictionary. Conducted a complete refactoring of user custom dictionary functionality including its HTTP API command “user_dictionary”. Its behavior and validation mechanisms now are adapted to a company-level or global custom dictionary. 
  • Input validation for newly added words and dictionary name. For example, words with special characters, punctuation marks or spaces can no longer be added to the dictionary. 
  • Updated error messages to be more precise and clear. Validated error codes returned for API responses.
  • Restricted input to a single word for addword and deleteword actions with word parameter. In upcoming releases, wordlist parameter will be added for actions with many words. 
• Núcleo de WProofreader:
  • Added UI localization for Indonesian language “id”. It can be enabled by adding the “localization: id” option to the config script.
  • Updated Spanish localization “es” for “Autocomplete suggestions” to ”Texto predictivo”.
• Other:
  • WProofreader SDK demos. Added a new demo with the WProofreader integration into TinyMCE v6.

Corrección de errores

• [WP-4593] Spelling: The word ”pay” is flagged as a misspelling in AmE
• [WP-4543] Spelling: Capitalized forms of British words are present in the AmE dictionary (e.g. Colour, Fibre, Paediatric, Aesthetic)
• [WP-4495] WProofreader core: isUserDictionaryExist method doesn’t corresponds its name
• [WP-4564] User dictionary HTTP API: Incorrect status code when “word” isn’t deleted
• [WP-4603] User dictionary HTTP API: Action to delete non-existent dictionary returns 200 OK
• [WP-4539] HTTP API: Invalid encoded URI returns BAD URI syntax
• [WP-4396] Specific words can’t be deleted using exclude functionality

Enhancement

AppServer

• Updated Hunspell spell check engine to version 1.7.1.
• Updated POCO C++ library to version 1.11.5.

Corrección de errores

• [WP-4557] HTTP API: The same check requests with enabled custom dictionary for American and British English returns different suggestions [Cloud product version]
• [WP-4566] HTTP API: Сheck request with custom_dictionary parameter returns error related to unsupported language Id [Cloud product version]*

*This fix will be applied on December 27, 2022 once the involved component is updated.

Hot fix release for the issue with AI requests that randomly lead to the application crash.

Note! We had to roll back the ICU changes made in v5.26.0 due to the issues that led to the application crashes causing 5xx errors. This change is applied in both cloud and on-premise product versions.

• Replaced sentence tokenization mechanism. The old custom sentence tokenization mechanism was replaced with the ICU word break iterator tool to improve processing of words that contain dots (abbreviations), hyphens, apostrophes, special symbols, shortening with one dot, possessions, etc. This approach is implemented for all languages that are based on the Hunspell spelling check engine.

Corrección de errores

• [WP-4180] Spelling check engine: Incorrect processing of words with trailing hyphens in German

Note! Only the on-premise product version was fully released. The cloud one is still in progress. We had to roll back the cloud release, specifically changes in the backend component, due to the issues that led to the application crashes causing 5xx errors. However, all the changes introduced in the WProofreader core are available in the cloud version as well.

Mejoras

Updated WProofreader core v3.8.4782:

• Global badge v2. Rethought the conception and behavior of the global badge. This badge no longer is tied to text fields but is global on the page. By default, it’s disabled globalBadge:false but if enabled it will be placed in the bottom right corner of the page. Its position can be regulated using the following options: badgeOffsetX, badgeOffsetY, badgeZIndex (by default, badgeOffsetX = 20, badgeOffsetY = 20, badgeZIndex = 1000).  

Grammar engine. Updated 3rd party engine, LanguageTool, to version 5.9.0.

• Added and improved grammar rules for the next languages: Catalan, Dutch, English, German, French, Polish, Portuguese, Russian, Spanish, Ukrainian.
• Updated 3rd party components to fix vulnerabilities in .jar libraries: CVE-2022-42003, CVE-2022-42004, CVE-2022-3171, CVE-2022-42889. 

Replaced sentence tokenization mechanism.

The old custom sentence tokenization mechanism was replaced with the ICU word break iterator tool to improve processing of words that contain dots (abbreviations), hyphens, apostrophes, special symbols, shortening with one dot, possessions, etc. This approach is implemented for all languages that are based on the Hunspell spelling check engine.

Corrección de errores

• [WP-4447] WProofreader core: No markup when user opens Full screen view in Google Gmail [detected using browser extension]
• [WP-4477] WProofreader core: No flag icon for Valencian (cat_ES) language
• [WP-4476] WProofreader core: Apostrophe (‘) is not allowed symbol for adding to the dictionary from UI
• [WP-4484] WProofreader core: WProofreader doesn’t work properly in IE11 in textareas
• [WP-4180] Spelling check engine: Incorrect processing of words with trailing hyphens in German
• [WP-4519] HTTP API: If set to “true”, ignore_mixed_case also affects and enables ignore_all_caps
• [WP-4501] Dictionaries: Some medical terms are recognized as incorrect words (e.g. AAOMR, ABthrax, ACTHoma, ALternaGEL)
• [WP-4491, WP-4524] Spelling: Words “filter”, “practice” and “groin” are flagged as misspellings in British English
• [WP-4452] AI engine: Misspelled words in quotes are recognized and underlined as grammar mistakes

Cambios de última hora

[On-premise product version] Updated the application server (AppServer) configuration file, AppServer.X.xml:

• Updated existing tags as follows:
  • Replaced Enabled tag for SpellCheck to Enabled attribute: <SpellCheck Enabled="true">
  • Added boolean Enabled attribute for Model, GrammarCheck, Autocomplete and Autocorrect tags.
• Added MaxInputSize parameter which is responsible for setting the max number of input tokens for AI language models in AppServer.X.xml. It adds flexibility in controlling model input size.

Cambios de última hora

[On-premise product version] Updated the application server (AppServer) configuration file, AppServer.X.xml:

• Added a new parameter / tag called CheckKit. It allows flexibly configuring the language settings.
• Renamed a list of the existing tags in the AppServer.X.xml 
  • GrammarCheckProvider → GrammarCheck
  • SpellEngineOptions → SpellCheck
  • FullPath → Path (it can work in both sides with a absolute path and relative path)
  • SpellCheckEnginePriority → SpellCheckPriority
  • AutoCompleteCheckProviderOptions → Autocomplete + <Locale>en</Locale>
  • AutoCorrect → Autocorrect

Mejoras

• BETA. English dialects support AI-based English under the hood. Implemented for next dialects:  American, British, Canadian and Australian English. By default it’s disabled but you can enable it as follows:
  • HTTP API: add “enforce_ai=true” to a query string;
  • WProofreader SDK integration: add “enforceAI: true” option to the configuration script.
• New version of WProofreader core v3.7.4760: 
  • Added option/parameter to turn ON/OFF enhanced text checking powered by AI.
  • Updated the default path for the WProofreader core for the cloud product version from “servicePath”: “spellcheck31/script/ssrv.cgi” to “servicePath”: “api”.
• Released v2.2.0 of the WProofreader plugin for CKEditor 5.  Added support of the Restricted editing mode.
• Improved AI-based English proofreading mechanism. Better handling cases with hyphenated compound words (e.g. mother-in-law), various typographical symbols and punctuation, contraction of verbs (e.g. “it’s”).    
• [On-premise product version] Updated the demo samples inside the package  to match the online version. By default the path to access the demos as follows: http: localhost/wscservice/samples/

Corrección de errores

• [WP-4399] WProofreader core: Styles were broken in the title of the JIRA issue [detected using browser extension]
• [WP-4428] WProofreader core: Disabled autocorrect feature for Slate.js
• [WP-4418] WProofreader core: Focus state shifted in Google Chat  [detected using browser extension]
• [WP-4405] AppServer: “System exception“ error occurred during license activation
• [WP-4442] [Cloud] BrE spelling: “tired” word marked as misspelled
• [WP-4451] [Cloud] BrE spelling: “benifitting”/”benifitted” words are marked as misspelled
• [WP-4443] AI engine: Incorrect handling suggestions to errors with zero length
• [WP-4411] HTTP API: Checking text with “%” causes BAD URI syntax error

Mejoras

Updated WProofreader core v3.6.4736

• Changed the badge behavior and view. By default, it starts in a small size and has a few states: orange – all is OK, no suggestions found; red – some suggestions found. On hover, users can check the exact number of alerts/suggestions detected. To return it to a large size use the option fullSizeBadge: true.
• Implemented the 1st version of the global badge under globalBadge option name. If enabled, there will be only one global badge on the page that aggregates all the suggestions on the page instead of having separate badges for each field.  By default, it’s disabled: globalBadge: false.
• Changed the default state of the suggestion popup view by enabling additional actions like: proofread in dialog, settings, toggle.
• Deprecated the badge pulsation effect and related disableBadgePulsing​​ option. Also, with the introduction of the small badge, removed the mechanism with the dynamic badge-size depending on the content and the text field.
• Learn more in the updated version of WProofreader API documentation.

Idiomas

Updated the sources of the Hunspell dictionaries for the next languages:

• Australian English. Source: en_AU Hunspell Dictionary by Kevin Atkinson
• Canadian English. Source: en_CA Hunspell Dictionary by Kevin Atkinson
• American English. Source: en_US Hunspell Dictionary by Kevin Atkinson
• English South Africa. Source: en_ZA Hunspell Dictionary by Dwayne Bailey
• English New Zealand. Source: Firefox Extensions > New Zealand English Dictionary by Stephen
• British English. Source: Sourceforge>latest SCOWL dictionary
• Portuguese. Source: pt_PT Hunspell Dictionary by Tiago F. Santos
• Brazilian Portuguese. Source: pt_BR Hunspell Dictionary by Raimundo Moura

Otros

• [On-Premise product version] HTTP API demo added to the demo samples, the same as we have in our public demos.

Corrección de errores

• [WP-4384,GitHub#32] WProofreader plugin for CKEditor 5: Accessibility concerns related to WProofreader badge element
• [WP-4388] WProofreader core: Autocorrect messes up corrected words
• [WP-4392] Spelling: Portuguese and Brazilian Portuguese don’t handle words with “ções”and “não” ending

Cambios de última hora

[On-premise product version] Updated Docker configuration:

• Settings (e.g. protocol, domain, web-port, etc.) can now be changed on the container run. More details in README.
• Application server (AppServer) logs are now linked to stdout, so they can be found in docker logs. Note that the license activation logs that are generated during image creation are stored in Logs/img_build_logs.tar.gz.
• config.ini and configSSL.ini were removed, all build parameters are now ARGs in the Dockerfile and also can be set as ENV variables.

Mejoras

Languages/Engines:

• AI-based German. Integrated version 2 of the AI German language model with improved processing quality, F0.5 score increased by 0.05 points. Now the model produces fewer false positive results (incorrect suggestions).
• Grammar engine. Updated 3rd party engine, LanguageTool, to version 5.8.0. 
  • Added and improved grammar rules for the next languages: Catalan, English, German, French, Polish, Portuguese, Russian, Spanish, Ukrainian.
  • Updated 3rd party component to fix vulnerability: CVE 2022-25647. Removed vulnerable Jetty libraries from the package. 
• Dictionaries. Extended the German and English medical dictionaries with new terms reported by the clients (e.g. COVID-19, Moderna, AstraZeneca).

Custom dictionary API:

• Words added to custom dictionaries can be divided by next separators: \n, Space, Tab, comma.
• Formed the list of supported characters that can be added to the dictionary (letters, digits and some punctuation marks). Words containing only digits or punctuation as well as their combinations are not supported.

Corrección de errores

• [WP-4349, GitHub #11951] WProofreader core: The caret moves to a different block when pressing space after a word with grammar error in CKEditor 5
• [WP-4352] WProofreader core: Incorrect behavior of cursor restore during autocorrect in Draft.js

Cambios de última hora

[On-premise product version] Updated the automated install script:

• Updated web_server_type values : 1 – Apache HTTP Server, 2 – NGINX, 3 – Other
• Renamed web server related parameters:
  • restart_apache → restart_web_server
  • apache_config_dir → web_server_config_dir

Nota. Tanto la versión en la nube como la versión local del producto salen a la venta. La versión en la nube incluye todo lo de v5.19.1 y el motor de comprobación gramatical actualizado a la versión 5.7.

Mejoras

• Motorgramatical. Actualizado el motor de terceros, LanguageTool, a la versión 5.7. Añadidas y mejoradas las reglas gramaticales para los siguientes idiomas: Catalán, francés, alemán, polaco, portugués, ruso, español y ucraniano. También incluye versiones actualizadas de las dependencias #6423 y corrección del problema de multihilo #6423 en la v5.6.

Nota. Sólo se ha lanzado la versión del producto on-premise. La versión del producto en la nube aún está en proceso. Tuvimos que posponer el lanzamiento debido a la agresiva invasión militar de Rusia en Ucrania (WebSpellChecker está hecho por ucranianos y en Ucrania).

Mejoras

• Sugerencias de autocompletar (beta). Se ha implementado y añadido una nueva funcionalidad a los productos basados en WProofreader: sugerencias de autocompletar mientras se escribe en inglés. Sólo sugiere una palabra o frase corta, la sugerencia puede ser aceptada usando la flecha izquierda ">" en el teclado. En el futuro se ampliará a frases más completas u oraciones comunes. La función está desactivada por defecto, puede ser fácilmente activada desde la interfaz de usuario por los usuarios finales o añadida por el desarrollador a la configuración utilizando la opción de autocompletar.
• Motorgramatical. Actualizado el motor de terceros, LanguageTool, a la versión 5.6. Añadidas y mejoradas las reglas gramaticales para los siguientes idiomas: Alemán, catalán, español, francés, inglés, portugués, ruso y ucraniano.
• Mejoras de seguridad. Actualizado el motor gramatical de LanguageTool con todas sus dependencias para eliminar vulnerabilidades comunes. 
• Núcleo de WProofreader. La versión se actualiza a la v3.5.xxx con el lanzamiento de las sugerencias de autocompletar. Desactivado el efecto de pulsación de la insignia: "disableBadgePulsing: true".

Corrección de errores

• [WP-4176] El recuento de palabras es incorrecto en las estadísticas de uso de la versión del producto en la nube
• [WP-4181] Núcleo de WProofreader: Estilos CSS rotos en un entorno de cliente
• [WP-4171] Núcleo de WProofreader: El marcado falla en texareas con más de 65 535 caracteres [Chrome]

Mejoras

• [Nube] Acelerar el proceso de activación del servicio tras la suscripción. Antes, la suscripción tardaba entre 5 y 10 minutos en activarse. Ahora el servicio comienza a funcionar de inmediato.
• Habilitar diccionarios médicos en inglés, español y francés para los idiomas correspondientes y sus dialectos. 
• Activar la prioridad del motor de corrección ortográfica basado en el diccionario para el español y el alemán basado en la IA. 

Cambios de última hora

• La lista de idiomas admitidos se redujo de más de 160 a aproximadamente más de 80 idiomas. Sólo se han eliminado los idiomas impopulares y no utilizados. Sin embargo, los clientes que utilizan las versiones locales del producto tienen la opción de añadir manualmente los idiomas que faltan (motor de corrección ortográfica Hunspell). La lista actualizada de idiomas soportados.
• Se ha retirado el límite del número de idiomas que se pueden utilizar en las versiones del producto en la nube o en las instalaciones. El paquete de la versión local se entrega con todos los idiomas admitidos. Sin embargo, algunos de ellos están en estado desactivado. Se pueden habilitar fácilmente siguiendo las instrucciones de la siguiente guía. 

Corrección de errores

• [WP-4140] Las palabras con apóstrofe no se pueden añadir al diccionario del usuario

Mejoras

• Se introduce un nuevo punto de entrada de servicio para los servicios basados en la nube. Es aplicable a las llamadas directas a la API y a los componentes basados en la interfaz de usuario, como el complemento WProofreader. Para la compatibilidad con versiones anteriores, la forma antigua también funciona. 

Forma antigua: https://svc.webspellchecker.net/spellcheck31/script/ssrv.fcgi?cmd=[comando]&[parámetro]=[valor]

Nueva forma: https://svc.webspellchecker.net/api?cmd=[comando]&[parámetro]=[valor]