Describe the bug
When setting up a simulation with the SYNERGY dataset Leenaars_2019, ASReview gives the following error. I highly suspect there are more SYNERGY datasets that have the same issue: for a duplicate (as defined by the new grouping system) there are conflicting labels. I propose that within a group we can simply take the True value and override the uploaded data (the title and abstract are the same, so the label should be too). But, in the export it should probably retain the original (False) label.
This will also be an issue for users uploading their own labelled data, as these can contain these mismatches on purpose: maybe they only include the journal publication, not the preprint (even though the title and abstract are the same).
Traceback (most recent call last):
File "/Users/___/Documents/venvs/asreview-2/lib/python3.13/site-packages/asreview/webapp/_task_manager/task_manager.py", line 73, in run
self.func(*self.args)
~~~~~~~~~^^^^^^^^^^^^
File "/Users/___/Documents/venvs/asreview-2/lib/python3.13/site-packages/asreview/webapp/_tasks.py", line 31, in run_task
run_simulation(project)
~~~~~~~~~~~~~~^^^^^^^^^
File "/Users/___/Documents/venvs/asreview-2/lib/python3.13/site-packages/asreview/webapp/_tasks.py", line 104, in run_simulation
sim = Simulate(
project.db.input.get_df(),
...<3 lines>...
groups=project.db.input.get_groups(),
)
File "/Users/___/Documents/venvs/asreview-2/lib/python3.13/site-packages/asreview/simulation/simulate.py", line 210, in __init__
raise ValueError(
f"Groups should not contain conflicting labels: {e}"
) from e
ValueError: Groups should not contain conflicting labels: Group 1648 contains conflicting labels.
ERROR:asreview.task_manager:Failed to train model for project 023aee4779074391b2c846514f2f1362
To Reproduce
Create a simulation with the SYNERGY dataset Leenaars_2019
Version information
- OS: MacOS
- Browser: Safari
- ASReview version: 3.0.5
Describe the bug
When setting up a simulation with the SYNERGY dataset
Leenaars_2019, ASReview gives the following error. I highly suspect there are more SYNERGY datasets that have the same issue: for a duplicate (as defined by the new grouping system) there are conflicting labels. I propose that within a group we can simply take the True value and override the uploaded data (the title and abstract are the same, so the label should be too). But, in the export it should probably retain the original (False) label.This will also be an issue for users uploading their own labelled data, as these can contain these mismatches on purpose: maybe they only include the journal publication, not the preprint (even though the title and abstract are the same).
To Reproduce
Create a simulation with the SYNERGY dataset
Leenaars_2019Version information