Composition of Conditional Random Fields for Transfer Learning

Abstract

Many learning tasks have subtasks for which much training data exists. Therefore, we want to transfer learning from the old, general-purpose subtask to a more specific new task, for which there is often less data. While work in transfer learning often considers how the old task should affect learning on the new task, in this paper we show that it helps to take into account how the new task affects the old. Specifically, we perform joint decoding of separately-trained sequence models, preserving uncertainty between the tasks and allowing information from the new task to affect predictions on the old task. On two standard text data sets, we show that joint decoding outperforms cascaded decoding.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2005
Accession Number
ADA440381

Entities

People

  • Andrew McCallum
  • Charles Sutton

Organizations

  • University of Massachusetts Amherst

Tags

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Artificial Intelligence Software
  • Bayesian Networks
  • Computer Languages
  • Computer Science
  • Data Sets
  • Decoding
  • Electronic Mail
  • Language
  • Learning
  • Machine Learning
  • Models
  • Named Entity Recognition
  • Natural Language Processing
  • Probabilistic Models
  • Probability
  • Recognition

Fields of Study

  • Computer science

Readers

  • Computer Programming and Software Development.
  • Instructional Design and Training Evaluation.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks