Dynamic Tag Tables
When a tag group finishes scoring on a dataset, its scores are available as SQL views. Explore and analyze them in the Queries tab. To check whether scoring is complete, see Check Scoring Progress.
Dynamic tag views
Every dynamic tag group has four views, one for each level. They appear in the output of SHOW TABLES.
Replace [dynamic tag group name] with your group’s name in lowercase, with spaces replaced by underscores. For example, a group named Casual Social Celebrations has the view dt_casual_social_celebrations_video.
In every view, tag_name is stored in lowercase (for example, pickup trucks), even when the UI shows title case. Filter on the lowercase value.
Threshold sources
Each view has a threshold_decision column that is 1 when the tag applies and 0 when it doesn’t. The threshold it uses depends on the level:
To apply a different cutoff in SQL, filter on score instead of threshold_decision.
dt_[dynamic tag group name]_video
Join to coactive_table_video on coactive_video_id for metadata.
dt_[dynamic tag group name]_shot
This view has no video ID or timestamps. Join coactive_shot_id to coactive_table_slice.id to get video_id, start_time_ms, and end_time_ms, then join video_id to coactive_table_video.coactive_video_id for metadata.
dt_[dynamic tag group name]_visual
This view is keyframe-level and has no video ID. Join to coactive_table on coactive_image_id to get coactive_video_id and metadata. The view reflects the latest version of each tag.
dt_[dynamic tag group name]_transcript
Join coactive_transcript_id to coactive_table_transcription.id to get the segment’s video_id, text, start_time, and end_time.
Join for advanced analytics
Join the dynamic tag views with the core tables to aggregate scores by video, filter by keyframe timestamp (keyframe_time_ms), or add file metadata to your results. For the columns in each core table, see SQL Table Reference.
Example query
This query finds keyframes where the basketball tag applies and adds show metadata from each keyframe’s video. The metadata column is a map<string,string>, so read each value with metadata['key']. Keys depend on the metadata ingested with your videos. To list them, see Read metadata values, then replace SeriesName and EpisodeName with your own keys.
For more query patterns, such as tag distribution and the top tags for each image or video, see Analyzing Dynamic Tags with SQL.
