We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Temporal Effects on Pre-trained Models for Language Processing Tasks

Abstract: Keeping the performance of language technologies optimal as time passes is of great practical interest. Here we survey prior work concerned with the effect of time on system performance, establishing more nuanced terminology for discussing the topic and proper experimental design to support solid conclusions about the observed phenomena. We present a set of experiments with systems powered by large neural pretrained representations for English to demonstrate that {\em temporal model deterioration} is not as big a concern, with some models in fact improving when tested on data drawn from a later time period. It is however the case that {\em temporal domain adaptation} is beneficial, with better performance for a given time period possible when the system is trained on temporally more recent data. Our experiments reveal that the distinctions between temporal model deterioration and temporal domain adaptation becomes salient for systems built upon pretrained representations. Finally we examine the efficacy of two approaches for temporal domain adaptation without human annotations on new data, with self-labeling proving to be superior to continual pre-training. Notably, for named entity recognition, self-labeling leads to better temporal adaptation than human annotation.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2111.12790 [cs.CL]
  (or arXiv:2111.12790v1 [cs.CL] for this version)

Submission history

From: Oshin Agarwal [view email]
[v1] Wed, 24 Nov 2021 20:44:12 GMT (170kb,D)

Link back to: arXiv, form interface, contact.