Skip to content

Code & Data for the Paper "Time Masking for Temporal Language Models", WSDM 2022

License

Notifications You must be signed in to change notification settings

guyrosin/tempobert

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Time Masking for Temporal Language Models

This repository provides a reference implementation of the paper:

Time Masking for Temporal Language Models
Guy D. Rosin, Ido Guy, and Kira Radinsky
Accepted to WSDM 2022
Preprint: https://arxiv.org/abs/2110.06366

Abstract:

Our world is constantly evolving, and so is the content on the web. Consequently, our languages, often said to mirror the world, are dynamic in nature. However, most current contextual language models are static and cannot adapt to changes over time.
In this work, we propose a temporal contextual language model called TempoBERT, which uses time as an additional context of texts. Our technique is based on modifying texts with temporal information and performing time masking - specific masking for the supplementary time information.
We leverage our approach for the tasks of semantic change detection and sentence time prediction, experimenting on diverse datasets in terms of time, size, genre, and language. Our extensive evaluation shows that both tasks benefit from exploiting time masking.

Prerequisites

  • Create an Anaconda environment with Python 3.8 and install requirements:

      conda create -n tempobert python=3.8
      conda activate tempobert
      conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c nvidia
    
  • Obtain datasets for training and evaluation:

Usage

  • Train TempoBERT using train_tempobert.py. This script is similar to Hugging Face's language modeling training script (link), and introduces two new arguments: time_embedding_type, that is set to "prepend_token" by default, and time_mlm_probability, which can used to set a custom probability for time masking.
  • Evaluate TempoBERT using semantic_change_detection.py for semantic change detection and sentence_time_prediction.py for sentence time prediction.

Pointers

  • The modification to the input texts is performed in tokenization_utils_fast.py, in TempoPreTrainedTokenizerFast._batch_encode_plus().
  • Time masking is performed in temporal_data_collator.py.

About

Code & Data for the Paper "Time Masking for Temporal Language Models", WSDM 2022

Resources

License

Stars

Watchers

Forks

Languages