Active Learning Based on Transfer Learning Techniques for Text Classification

Abstract

Text preprocessing is a common task in machine learning applications that involves handlabeling sets. Although automatic and semi-automatic annotation of text data is a growing field, researchers need to develop models that use resources as efficiently as possible for a learning task. The goal of this work was to learn faster with fewer resources. In this paper, the combination of active and transfer learning was examined with the purpose of developing an effective text categorization method. These two forms of learning have proven their efficiency and capacity to train correct models with substantially less training data. We considered three types of criteria for selecting training points: random selection, uncertainty sampling criterion and active transfer selection. Experimental evaluation was performed on five data sets from different domains. The findings of the experiments suggest that by combining active and transfer learning, the algorithm performs better with fewer labels than random selection of training points. INDEX TERMS active learning, active transfer learning, text classification, transfer learning I. INTRODUCTION

Transfer Learning-Based Center-of-Mass Positioning Methods for Cultural Relics

“…To address the problem of scarce training data, Gikunda [21] et al proposed a model parameter migration framework, which provides near-optimal training capabilities for classification models using only a small set of samples through a heuristic combination of migration learning and deep learning methods. Onita [22] investigates the combination of active and transfer learning for text categorization, using three training point selection criteria to complete the learning task as efficiently as possible using fewer resources. Garcia [23] et al proposed a multi-source task migration learning method to address the problem of high time cost of model training by applying a correlation convolution kernel to match a target model with a similar number of parameters as the source model and reducing the cost of source model training by pretraining the target model for parameter migration.…”

A Novel Levy Walk-Based Framework for Scheduling Power-Intensive Mobile Edge Computing Tasks

“…CNNs are well suited for multi-task learning, in which a single model is trained to carry out several related applications concurrently [162]- [177]. The need for large amounts of labeled data for each individual task is being reduced as researchers investigate ways to take advantage of shared representations across applications and enhance generalization by transferring knowledge learned from one task to another [147]- [161].…”

Impacto del preprocesamiento en la clasificación automática de textos usando aprendizaje supervisado y reuters 21578

“…No obstante, etiquetar estos contenidos manualmente requiere de considerable esfuerzo humano y tiempo. Aquí es donde la tecnología contemporánea cobra relevancia, impulsando el desarrollo de herramientas y programas que, mediante la inteligencia artificial, automatizan este proceso [2], [3], [4], [5], [6].…”