Autocomplete suggestions automatically complete the user’s thought by suggesting the next word or a short phrase based on the context. 

To download it for the on-premise version you must have the Internet connection enabled and have free space available on the disk according to Hardware requirements for v5.7.0+. Also, the English autocomplete element has to be chosen while installation. The Autocomplete language models will be downloaded automatically if these criteria are met. Otherwise, the download will fail and you can download the models manually and enable them accordingly.

  1. Contact technical support and request links to download Autocomplete language models.
  2. Download the language models for required languages to the WebSpellChecker\AppServer\autocomplete\neural\ directory.
  3. Unpack the downloaded models.
Example of the Linux unpack command for English model
tar -xvzf en.tar.gz

      4. Go to AppServerX.xml and navigate to the language tags.

      5. Сhange the value from false to true for the Enabled tag. 

<AIModels>
			<Model Enabled="true" Lang="en" ModelName="en" Type="grammar" WarmUpId="en_AI" SessionsNumber="1"/>
			<Model Enabled="true" Lang="de" ModelName="de" Type="grammar" WarmUpId="de_AI" SessionsNumber="1"/>
			<Model Enabled="true" Lang="es" ModelName="de_es" Type="grammar" WarmUpId="es_AI" SessionsNumber="1"/>
			<Model Enabled="true" Lang="en" ModelName="en" Type="autocomplete" WarmUpId="en_AI" ProbabilityThreshold="0.15" MaxInputSize="200" SessionsNumber="1"/>
</AIModels>

For the WebSpellChecker version older than 5.26.0.0, in point #5 you just need to uncomment the <AutoCompleteCheckProviderOptions>en</AutoCompleteCheckProviderOptions> tag in the AppServerX.xml file. This is relevant for en_AI, en_US, en_GB, en_CA, en_AU, en_NZ, en_ZA languages.

      6. Restart AppServer to apply changes.

If you have trouble with the described steps, it's better to use the automated option while the initial installation when an Internet connection is in place.

  • No labels