topics-extractor

To extract topics with timestamps, the Skill performs semantic matching against a meeting summary to map each topic to a time range. It collects anchor quotes and speaker information, outputting a structured JSON file with start and end seconds.

Aligning summary topics with transcript time ranges is done through semantic matching to produce a topics JSON file. This file ensures times are accurate and non-overlapping, capturing id, title, startSec, endSec, anchorTexts, speakers, and importance.

Yes, a meeting summary is required to detect topics in a transcript. The extraction uses semantic matching against the summary to identify topic segments, collect anchor quotes, and map them to accurate start and end times.

The extracted topic data uses a JSON format saved to meeting_rec/transcript/topics.json. Each topic entry contains id, title, summary, startSec, endSec, anchorTexts, speakers, and an importance rating.

To ensure extracted topic timestamps are non-overlapping, the semantic matching process calculates accurate time ranges. It outputs the results in a structured JSON file, verifying that the startSec and endSec values do not overlap where applicable.