Note that first the data for these models should be provided and processed (as described in data/) and then each model can be trained individually.
If OOM occurs, we suggest that reducing train_batch_size and increasing gradient_accumulation_steps (gradient_accumulation_steps is used to perform Gradient Accumulation).
Inference
After training and saving the Constituent Extraction and Constituent Linking models, test the pipelined CompactIE on evaluation benchmarks. Three popular evaluation benchmarks (BenchIE, CaRB, Wire57) are provided to examine CompactIE's performance.
python test.py --config_file config.yml
To test the CompactIE system on a set of sentences, first process those sentences as described here. Then, modify the config.yml file so that the path to the processed sentence file (test_file) and its conjunction file (conjunctions_file) is properly reflected. Finally, run the above command to produce extractions.
@inproceedings{fatahi-bayat-etal-2022-compactie,
title = "{C}ompact{IE}: Compact Facts in Open Information Extraction",
author = "Fatahi Bayat, Farima and
Bhutani, Nikita and
Jagadish, H.",
booktitle = "Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
month = jul,
year = "2022",
address = "Seattle, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.naacl-main.65",
doi = "10.18653/v1/2022.naacl-main.65",
pages = "900--910",
abstract = "A major drawback of modern neural OpenIE systems and benchmarks is that they prioritize high coverage of information in extractions over compactness of their constituents. This severely limits the usefulness of OpenIE extractions in many downstream tasks. The utility of extractions can be improved if extractions are compact and share constituents. To this end, we study the problem of identifying compact extractions with neural-based methods. We propose CompactIE, an OpenIE system that uses a novel pipelined approach to produce compact extractions with overlapping constituents. It first detects constituents of the extractions and then links them to build extractions. We train our system on compact extractions obtained by processing existing benchmarks. Our experiments on CaRB and Wire57 datasets indicate that CompactIE finds 1.5x-2x more compact extractions than previous systems, with high precision, establishing a new state-of-the-art performance in OpenIE.",
}
Contact
In case of any issues or question, please send an email to farimaf (at) umich (dot) edu.
