Language support
IBM Watson® Knowledge Studio provides support for training a model in several languages.
Support for multiple languages does not mean that the product itself is translated. The Knowledge Studio user interfaces, messages, and documentation are available in English only.
You can train a model in the following languages:
- English (the default language)
- Arabic
- Brazilian Portuguese
- Chinese (Simplified)
- Chinese (Traditional)
- Dutch
- French
- German
- Italian
- Japanese
- Korean
- Spanish
Support includes the ability to add documents in these languages to a workspace, add dictionaries, run pre-annotation, use the ground truth editor to annotate documents, and train a machine learning model. When you select a language, the system applies language-specific templates to handle dictionary entries, text tokenization, and sentence segmentation. For information about how the system handles dictionaries in different languages, see Dictionaries.
The rule editor and rule-based model cannot handle bidirectional text, so cannot be used with Arabic documents.
Restrictions
Because of character restrictions in the underlying machine learning technology, entries in the type system cannot contain spaces and can include only the following ASCII characters:
- A through Z and a through z
- 0 through 9
- The underscore character: _
The following rules also apply:
- The first character in the name must be alphabetical.
- The entity type name length cannot exceed 64 characters.
- The relation type name length cannot exceed 128 characters.
Thus, for example, when a human annotator annotates Japanese text in the ground truth editor, the entity type and relation type names that can be applied will not be in Japanese.