/nocode-er-bench

No-code Benchmarking of Entity Resolution

Primary LanguagePython

No-code Benchmarking of Entity Resolution

https://swimlanes.io/u/k3Rmy375P

This is great news! The main requirement is to implement the following process: to set several DL-based matching algorithms running on a server as Docker containers to feed one of them with three sets of record pairs from pyJedAI' blocking (training, validation and testing), to receive the labels of the pairs in the testing set

We will also need some visualizations, but this should be easy. The docker images are available here: https://github.com/gpapadis/DLMatchers/tree/main/dockers/mostmatchers The scenario is described in more detail in Section 3.2 here: https://www.overleaf.com/9862813521fjcqwzqnbmxc pyJedAI is available here: https://github.com/AI-team-UoA/pyJedAI

I forgot to mention that the dataset is here: https://github.com/gpapadis/DLMatchers/tree/main/EMTransformer/data/abt_buy . The model uses the files train.csv, test.csv and valid.csv.

To build container, run

docker build -t emtransformer emtransformer docker run -it --entrypoint=/bin/bash --gpus all emtransformer

Some more matcher scripts are available here: https://github.com/nishadi/DLMatchers/tree/main/dockers/mostmatchers/scripts