Signal Configuration

Keyword-Based Detection

Keyword-based detection matches specific words or phrases in transcripts:

Setting Up Keywords:

1. Select "Keyword" as your detection type
2. Enter keywords or phrases (one per line)
3. Choose match type:

• Exact Match: Matches the exact phrase
• Partial Match: Matches if the phrase appears anywhere in a sentence
• Word Boundary Match: Matches whole words only

4. Configure case sensitivity
5. Set minimum occurrence count (how many times it must appear)

Advanced Options:

• Exclusion Words: Add words that, when present, should exclude the match
• Context Requirements: Require certain words to appear nearby
• Multiple Keywords: Use AND/OR logic to combine multiple keywords

Example Configuration:

• Keywords: "security breach", "data leak", "unauthorized access"
• Match Type: Partial Match
• Case Sensitive: No
• Minimum Occurrences: 1

AI Prompt-Based Detection

AI prompt-based detection uses natural language understanding:

Writing Effective Prompts:

1. Clearly describe what you want to detect
2. Provide examples of what should be detected
3. Specify what should NOT be detected (to reduce false positives)
4. Include context requirements
5. Set confidence threshold

Prompt Structure:

• What to Detect: Clear description of the concept
• Examples: Positive examples of what should trigger the signal
• Exclusions: Negative examples of what should NOT trigger
• Context: Required surrounding context or conditions

Example Prompt:

Detect when customers express dissatisfaction or frustration with the product or service. 

This includes complaints about features, support quality, or overall experience.

Examples: "This feature doesn't work", "I'm frustrated with the support", "The product is too slow"

Exclude: General questions or feature requests without negative sentiment

Context: Must include emotional indicators or negative language

Detection Rules Setup

Configure advanced detection rules:

Rule Conditions:

• Multiple Conditions: Combine multiple detection criteria with AND/OR logic
• Time-Based Rules: Detect signals only during specific time periods
• Speaker-Based Rules: Detect signals only from specific speakers
• Frequency Rules: Require signals to appear multiple times before triggering

Rule Examples:

• Detect "budget approval" only if mentioned by a manager
• Detect "security incident" only if it appears in the first 10 minutes
• Detect "competitor mention" only if followed by a comparison

Signal Filters

Apply filters to refine signal detection:

• Domain Filters: Only detect signals in specific domains
• Recording Type Filters: Only scan certain types of recordings
• Meeting Type Filters: Only detect in specific meeting types
• Date Range Filters: Only detect signals in recordings from certain date ranges
• Folder Filters: Only scan recordings in specific folders

Filters help you focus signal detection on the most relevant content.

Signal Tags

Add tags to your custom signals for better organization:

• Create tags to group related signals
• Use tags for filtering and searching
• Examples: "urgent", "sales", "compliance", "customer-facing"
• Signals can have multiple tags

Tags make it easier to find and manage related signals.