Analytics
Last updated
Last updated
The Analytics page offers a detailed overview of how your systems are performing. It focuses on three key components: Utterances, Conversations, and Tags.
Utterances: Displays individual user inputs, allowing you to analyze how well the system recognizes and responds to these utterances. This section helps in understanding user intent, improving accuracy, and identifying any gaps in the training data.
Conversations: Provides an overview of full interactions between users and the system. By analyzing conversations, you can track patterns, identify common issues, and measure engagement to improve user experience.
Tags: Organizes and categorizes utterances based on pre-defined intents. This feature helps in managing large datasets, simplifying searches for specific interactions, and improving the overall organization of data for analysis and future enhancements.
By leveraging these tools, you can gain deep insights into user behavior, system performance, and areas for improvement, making your interactions more efficient and accurate.
You can now keep track of the confidence level in the match annotation in the conversations.
Confidence Score: Each utterance is assigned a confidence score, typically represented as a percentage. This score indicates how confident the system is that the recognized intent aligns with the user’s input. For example, a confidence level of 90% suggests high certainty that the detected intent matches the input.
Reviewing in Conversations: Within the conversation data, you can see the confidence levels associated with each utterance. This allows you to analyze whether the system is making accurate predictions and helps identify cases where it might need more training or adjustments.
Improving Confidence: If the confidence levels are consistently low for certain utterances, it indicates the need for refining training data, improving the intent definitions, or clarifying ambiguous user inputs to enhance system performance.
By monitoring and adjusting based on confidence levels in the match annotation, you can ensure that the system responds more accurately and reliably in conversations.
The Conversations section of the Analytics page provides a comprehensive view of interactions between users and your system. This tool helps in understanding user behavior, tracking system performance, and identifying areas for improvement.
View Complete Interactions: In the Conversations section, you can view full dialogue exchanges between users and the system. Each conversation is recorded step-by-step, including user inputs (utterances) and the system’s responses. This allows you to see the context and flow of the interaction in detail.
Monitor Accuracy and Performance: By reviewing conversations, you can analyze how well the system handles different user inputs. You’ll be able to see where the system performs correctly and where it might struggle, such as with misinterpreted utterances or incorrect intent matching.
Identify Patterns: As you explore multiple conversations, you can identify recurring issues, common questions, or frequently misunderstood intents. These insights can guide system improvements and help optimize the overall user experience.
By regularly reviewing and analyzing Conversations, you can enhance system accuracy, improve user interactions, and ensure the performance of your system aligns with user needs and expectations.
The Tags section in the Analytics page is a powerful tool for organizing and categorizing utterances and conversations, helping you analyze interactions more efficiently.
Categorizing Data: Tags allow you to assign specific labels to utterances and conversations based on their intent, topic, or any custom criteria. For example, you can create tags like “Support Request,” “FAQ,” or “Product Inquiry” to group similar interactions together. This helps in organizing large datasets for easier analysis.
Filtering Conversations and Utterances: You can filter the data using tags, enabling you to focus on specific types of interactions. This is especially useful when you want to track performance for a particular intent or examine recurring issues within a certain category.
Tracking Performance by Tag: Once you have categorized your interactions using tags, you can track how well the system performs for each category. This allows you to identify which tags are associated with high or low confidence scores, giving you insight into where the system excels or needs improvement.
Improving System Training: Tags make it easier to identify patterns and trends within specific categories of interactions. If you notice that certain tags are frequently associated with low confidence scores or misunderstood intents, you can use this information to refine the system’s training data and improve accuracy.
Customizing Tags: You have the flexibility to create custom tags that suit the specific needs of your system or business. These tags can reflect any level of granularity, such as broad categories like “General Queries” or more specific ones like “Billing Issues.”
By using Tags, you can enhance your analysis of user interactions, improve system organization, and make informed decisions to fine-tune performance and enhance the user experience.
