NLU Ontology and Semantic Networks
The Czech Lexical Resource has been under special care during this release and has been enhanced with + 9.000 Language Objects and Entities. The improvements include new Entities for feminine and masculine Czech names, where all names are conjugated and the variables
sNameVocative available. Diminutives, informal and familiar forms are covered for the majority of the names.
Lists of more than 5.000 conjugated Czech cities have also been added.
The following table summarizes the types of the new Language Objects. For a description of each Language Object type please visit the Pre-built Knowledge section.
|Language Object suffix||Count|
|LIST||General lists: 14 Emojis lists: 17 Names lists: 40 (covering world feminine and masculine names) Numerals lists: ordinal and cardinal numbers 1- 100, all numerals fully declined (covering also informal forms)|
On top of the new Language Objects, an elevated number of existing Language Objects have been reviewed and enhanced.
Read more about Teneo Lexical Resources
Czech Teneo Dialogue Resource
Teneo 7.3 brings a completely new Czech Teneo Dialogue Resource to the list of resources in the Teneo Platform.
The Czech Teneo Dialogue Resource consists of 100 Flows covering the generic dialogue capabilities that the Conversational AI application needs to behave human-like, show social skills and personalized behavior.
As with all the Teneo Dialogue Resources, the Czech TDR is nearly production ready, although some project-time should be dedicated to adapting the flows manually according to the needs of the project. The necessary adaptations may include ensuring that the TDR triggers and order groups are compatible with the project's triggers and order groups, or adapting bot outputs to follow corporate branding. Projects can also re-arrange and re-name folders and change the structure if needed.
Variable names in the Teneo Dialogue Resources have been changed to follow best practices and use camelCase; therefore, the names are updated from, for example,
botName or from
userFirstName. Note that global variables now includes the description Teneo Dialogue Resource variable instead of the prefix "Lib_" known from previous versions of the Dialogue Resources.
Flow variables in the horoscope flow, language detector flow and the favorites flow are also updated.
Predict Input Processor
The Predict Input Processor is extended to cater for the usage of CLU models in Teneo and Predict now supports both Learn and CLU intent models. With the implementation of deferred or lazy intent classification the Predict Input Processor now only creates annotations on demand when Engine calls the input processor specifically during the input matching process.
When the Predict Input Processor is called and Predict receives an input - as always - confidence scores are calculated for each class based on the model and annotations created for the most confident classes; a new
classifier variable is attached to the class annotations in Teneo 7.3 providing information related to the used classifier, and the value is CLU or Learn.
Note that, in case of network issues, Azure downtime or similar, a Learn model is always sent along which can be used in the mentioned scenarios as fallback and in that case the value of the
classifier variable is LearnFallback.
Input Processors Logging to Engine Logs
With the option to use CLU in Teneo, the Input Processors are extended to enable them to send log messages to the engine log message and the Engine will then log it together with the session Id and the transaction Id (if available). Apart from engine logs, these messages are also visible in Teneo Studio's Tryout and under IP Errors in the Publish Status of a Publish Environment.
Also see the Teneo Engine section.