Fine-tune Your Assistant Performance
with in-built testing tools
Interact with your virtual assistant in real-time before you publish – to test recognition, performance, and flow of the conversation for various scenarios. The XO Platform’s in-built tools enable you to try different variations of user prompts to ensure the NLP interpreter is accurately processing these utterances.
Testing virtual assistants with a good sample of utterances gives invaluable insights into the performance.
Understand how the Kore.ai XO Platform process the user request through its’ multi-engine. Check if it identifies the right intent and entities for a given utterance; fine-tune the training based on the expected response.
Delineate how well the virtual assistant understands user utterances using predefined test suites available in the builder or custom test suites. Execute a series of tests to get a detailed statistical analysis and gauge the performance of the ML model; view NLP metrics like precision, recall & F1 scores.
Use the Platform’s visualization tools, confusion matrix and K-fold testing to understand the assistant usage patterns, conversation flows and drop-off points.
Further, drill down results to intent and test case level analysis. Tag specific test case results that need follow-up actions and collaborate with your team to improve the performance.
Streamline virtual assistant updates confidently by creating conversation test cases for handling regression. You can now create a test by uploading a file. The Flow, Text, and Context assertions test entire flows and allow you to compare versions. The Flow coverage and performance analytics provide helpful insights.
NLU Test Data Suggestions
Batch Test suites help evaluate the NLU performance using test datasets. The new Generative AI capabilities analyze IVA scope and generate test data for various scenarios. These test suites help you achieve higher coverage by including unseen phrases, entity checks, spelling mistakes, and negative cases.
The Kore.ai XO Platform provides complete transparency into intent identification and entity extraction with explanations about each query response.
The debug log gives you info on what is happening at the backend – you see the NLP, logs, session context, and variables. Unlock the “black box” of machine learning and use the data to analyze the root cause of various failure scenarios.
Health & Monitoring
Actionable recommendations suggest essential changes to improve overall NLU and Flow health. The new upgrade includes the entire Flow Health framework. Drill down to specific intents to identify coverage and performance analytics and review uncovered paths to develop new tests.
Learn how Kore.ai products and solutions can drive ROI, reduce expenses and increase revenue.