Calibration, Entropy Rates, and Memory in Language Models

Mark Braverman, Xinyi Chen, Sham Kakade, Karthik Narasimhan, Cyril Zhang, Yi Zhang
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:1089-1099, 2020.

Abstract

Building accurate language models that capture meaningful long-term dependencies is a core challenge in natural language processing. Towards this end, we present a calibration-based approach to measure long-term discrepancies between a generative sequence model and the true distribution, and use these discrepancies to improve the model. Empirically, we show that state-of-the-art language models, including LSTMs and Transformers, are miscalibrated: the entropy rates of their generations drift dramatically upward over time. We then provide provable methods to mitigate this phenomenon. Furthermore, we show how this calibration-based approach can also be used to measure the amount of memory that language models use for prediction.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-braverman20a, title = {Calibration, Entropy Rates, and Memory in Language Models}, author = {Braverman, Mark and Chen, Xinyi and Kakade, Sham and Narasimhan, Karthik and Zhang, Cyril and Zhang, Yi}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {1089--1099}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/braverman20a/braverman20a.pdf}, url = {https://proceedings.mlr.press/v119/braverman20a.html}, abstract = {Building accurate language models that capture meaningful long-term dependencies is a core challenge in natural language processing. Towards this end, we present a calibration-based approach to measure long-term discrepancies between a generative sequence model and the true distribution, and use these discrepancies to improve the model. Empirically, we show that state-of-the-art language models, including LSTMs and Transformers, are miscalibrated: the entropy rates of their generations drift dramatically upward over time. We then provide provable methods to mitigate this phenomenon. Furthermore, we show how this calibration-based approach can also be used to measure the amount of memory that language models use for prediction.} }
Endnote
%0 Conference Paper %T Calibration, Entropy Rates, and Memory in Language Models %A Mark Braverman %A Xinyi Chen %A Sham Kakade %A Karthik Narasimhan %A Cyril Zhang %A Yi Zhang %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-braverman20a %I PMLR %P 1089--1099 %U https://proceedings.mlr.press/v119/braverman20a.html %V 119 %X Building accurate language models that capture meaningful long-term dependencies is a core challenge in natural language processing. Towards this end, we present a calibration-based approach to measure long-term discrepancies between a generative sequence model and the true distribution, and use these discrepancies to improve the model. Empirically, we show that state-of-the-art language models, including LSTMs and Transformers, are miscalibrated: the entropy rates of their generations drift dramatically upward over time. We then provide provable methods to mitigate this phenomenon. Furthermore, we show how this calibration-based approach can also be used to measure the amount of memory that language models use for prediction.
APA
Braverman, M., Chen, X., Kakade, S., Narasimhan, K., Zhang, C. & Zhang, Y.. (2020). Calibration, Entropy Rates, and Memory in Language Models. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:1089-1099 Available from https://proceedings.mlr.press/v119/braverman20a.html.

Related Material