CSV Format Requirements
Base Required Columns (All Modes)
| Column | Type | Description |
|---|---|---|
turn_id | integer | Unique identifier for each utterance |
speaker | string | Speaker identifier (e.g., "Teacher", "S1") |
start | float | Start time in seconds |
end | float | End time in seconds |
utterance | string | The spoken text |
Mode-Specific Columns
| Column | Mode | Description |
|---|---|---|
context_id | Context | Groups turns into contexts |
target | Context | 1 or true for the turn to annotate |
pair_id | Pair | Groups 2 turns for comparison |
batch_id | Batch | Groups turns into batches |
Example CSV (All Columns)
A single CSV file can include all mode-specific columns, allowing you to use the same file with different annotation modes:
turn_id,speaker,start,end,utterance,context_id,target,pair_id,batch_id 1,Teacher,0.0,2.5,"Hello class",1,,1,batch_001 2,Teacher,3.0,5.2,"Today's topic",1,1,1,batch_001 3,Student,6.1,8.9,"I have a question",2,,2,batch_001 4,Teacher,9.0,11.0,"Yes, go ahead",2,1,2,batch_002
Tips
- Time values (start, end) should be in seconds with decimal precision
- Use double quotes around utterances containing commas
- Empty cells are allowed for mode-specific columns you don't need
- The
targetcolumn accepts: 1, true, TRUE, or empty - Keep CSV files under 10 MB for best performance