IncDSI: Incrementally Updatable Document Retrieval

Varsha Kishore, Chao Wan, Justin Lovelace, Yoav Artzi, Kilian Q Weinberger
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:17122-17134, 2023.

Abstract

Differentiable Search Index is a recently proposed paradigm for document retrieval, that encodes information about a corpus of documents within the parameters of a neural network and directly maps queries to corresponding documents. These models have achieved state-of-the-art performances for document retrieval across many benchmarks. These kinds of models have a significant limitation: it is not easy to add new documents after a model is trained. We propose IncDSI, a method to add documents in real time (about 20-50ms per document), without retraining the model on the entire dataset (or even parts thereof). Instead we formulate the addition of documents as a constrained optimization problem that makes minimal changes to the network parameters. Although orders of magnitude faster, our approach is competitive with re-training the model on the whole dataset and enables the development of document retrieval systems that can be updated with new information in real-time. Our code for IncDSI is available at https://github.com/varshakishore/IncDSI.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-kishore23a, title = {{I}nc{DSI}: Incrementally Updatable Document Retrieval}, author = {Kishore, Varsha and Wan, Chao and Lovelace, Justin and Artzi, Yoav and Weinberger, Kilian Q}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {17122--17134}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/kishore23a/kishore23a.pdf}, url = {https://proceedings.mlr.press/v202/kishore23a.html}, abstract = {Differentiable Search Index is a recently proposed paradigm for document retrieval, that encodes information about a corpus of documents within the parameters of a neural network and directly maps queries to corresponding documents. These models have achieved state-of-the-art performances for document retrieval across many benchmarks. These kinds of models have a significant limitation: it is not easy to add new documents after a model is trained. We propose IncDSI, a method to add documents in real time (about 20-50ms per document), without retraining the model on the entire dataset (or even parts thereof). Instead we formulate the addition of documents as a constrained optimization problem that makes minimal changes to the network parameters. Although orders of magnitude faster, our approach is competitive with re-training the model on the whole dataset and enables the development of document retrieval systems that can be updated with new information in real-time. Our code for IncDSI is available at https://github.com/varshakishore/IncDSI.} }
Endnote
%0 Conference Paper %T IncDSI: Incrementally Updatable Document Retrieval %A Varsha Kishore %A Chao Wan %A Justin Lovelace %A Yoav Artzi %A Kilian Q Weinberger %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-kishore23a %I PMLR %P 17122--17134 %U https://proceedings.mlr.press/v202/kishore23a.html %V 202 %X Differentiable Search Index is a recently proposed paradigm for document retrieval, that encodes information about a corpus of documents within the parameters of a neural network and directly maps queries to corresponding documents. These models have achieved state-of-the-art performances for document retrieval across many benchmarks. These kinds of models have a significant limitation: it is not easy to add new documents after a model is trained. We propose IncDSI, a method to add documents in real time (about 20-50ms per document), without retraining the model on the entire dataset (or even parts thereof). Instead we formulate the addition of documents as a constrained optimization problem that makes minimal changes to the network parameters. Although orders of magnitude faster, our approach is competitive with re-training the model on the whole dataset and enables the development of document retrieval systems that can be updated with new information in real-time. Our code for IncDSI is available at https://github.com/varshakishore/IncDSI.
APA
Kishore, V., Wan, C., Lovelace, J., Artzi, Y. & Weinberger, K.Q.. (2023). IncDSI: Incrementally Updatable Document Retrieval. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:17122-17134 Available from https://proceedings.mlr.press/v202/kishore23a.html.

Related Material