Hi Ines,
I'm doing a ner.manual task and i feed in a csv dataset with 2 columns id, text. After annotating I exported the jsonl file and everythinh worked fine. I'd like to know if there is a possibility to get back the id field from _input_hash and _task_hash of the jsonl file.
In a lot of tasks I do, I feed in the same format (a csv with id,text columns) I'm wondering if there is a way to add the id field to the jsonl
Thank you
Hi! By default, the CSV loader will only read from the columns Text, Label and Meta. If you want to include other custom data and properties the easiest way is to use a JSON-based format like JSON or JSONL instead. This lets you represent nested structures and different data types, and all custom properties you add are passed through with the data. You could also write a custom loader that takes your CSV and sends out JSON data with any transforms and custom properties you need.
The input and task hashes are based on the existing task fields, like the "text". You can read more about them here.