I want: The system to analyze the meeting transcription and automatically determine the type of meeting,
So that: I can streamline the post-meeting analysis process and gain insights specific to the detected meeting type.
Acceptance Criteria
Transcription Analysis: Upon concluding the meeting, the system should evaluate the transcription against the predefined meeting types.
Model Selection: Utilize a lighter model, potentially less resource-intensive than GPT-3.5, to efficiently match the transcript to a meeting type.
Independent Information Extraction: Regardless of the detected meeting type, the system can also try to extract common information like action items from the transcription.
Confidence Levels: If we don't know what the meeting type was or if there is more than one potential match we will select the one with the highest likelihood. If we can't assess the type with confidence at least 90% we leave the type undefined. We can optionally give the user a list of our best guesses and let them choose from our suggestions.
Notes
While accuracy is crucial, ensuring speed in detection will enhance user experience.
AI might not be sure about the meeting type. A meeting can often confidently match to several types. We should set a threshold of confidence below which we don't preset the meeting type automatically.
Out of Scope
Detailed analysis specific to the detected type (this is handled by other features).
Dependencies
Comprehensive list of existing meeting types with associated characteristics.
Title: Detect Meeting Type from Transcription
As a: User of LUNI.AI
I want: The system to analyze the meeting transcription and automatically determine the type of meeting,
So that: I can streamline the post-meeting analysis process and gain insights specific to the detected meeting type.
Acceptance Criteria
Transcription Analysis: Upon concluding the meeting, the system should evaluate the transcription against the predefined meeting types.
Model Selection: Utilize a lighter model, potentially less resource-intensive than GPT-3.5, to efficiently match the transcript to a meeting type.
Independent Information Extraction: Regardless of the detected meeting type, the system can also try to extract common information like action items from the transcription.
Confidence Levels: If we don't know what the meeting type was or if there is more than one potential match we will select the one with the highest likelihood. If we can't assess the type with confidence at least 90% we leave the type undefined. We can optionally give the user a list of our best guesses and let them choose from our suggestions.
Notes
While accuracy is crucial, ensuring speed in detection will enhance user experience. AI might not be sure about the meeting type. A meeting can often confidently match to several types. We should set a threshold of confidence below which we don't preset the meeting type automatically.
Out of Scope
Detailed analysis specific to the detected type (this is handled by other features).
Dependencies
Comprehensive list of existing meeting types with associated characteristics.