Dirty ER examples input .csv

[vromaniello]

Hi, it is possible to have sample files in .csv format for

• entity profile D1
• ground truth
  because .csv files with any formatting will not work.
  The error from JedAI-gui is the following:

Thanks you for the support

[gpapadis]

Hi Vito,

thanks for your interest in JedAI!

The error you get is caused by an incompatibility between JedAI-gui and JedAI-core. The former is practically deprecated. It's better to use JedAI's web app through Docker. See Table 4 here for instructions: https://helios2.mi.parisdescartes.fr/~themisp/publications/is21-jedaiRepro.pdf .

Still, our custom implementation of the CSV reader gets confused by the separator, when it's not a comma. The best approach is to format your dataset like the Leipzig benchmarks (https://dbs.uni-leipzig.de/research/projects/object_matching/benchmark_datasets_for_entity_resolution), using comma as a separator and quotes for the values. Below you can find screenshots showing that JedAI can successfully read the csv files of the Abt-Buy dataset.

Btw, some of the Leipzig datasets involve more than than one duplicate per entity, even though they are Clean-Clean ER datasets. This is not supported by JedAI, which automatically removes equivalence clusters with more than two entities and prints the relevant messages in the command line.

[vromaniello]

Thanks! My tests succedeed with JedAI's web app through Docker.