Abstract
The purpose of document classification is to assign the most appropriate label to a specified document. The main challenges in document classification are insufficient label information and unstructured sparse format. A semi-supervised learning (SSL) approach could be an effective solution to the former problem, whereas the consideration of multiple document representation schemes can resolve the latter problem. Co-training is a popular SSL method that attempts to exploit various perspectives in terms of feature subsets for the same example. In this paper, we propose multi-co-training (MCT) for improving the performance of document classification. In order to increase the variety of feature sets for classification, we transform a document using three document representation methods: term frequency–inverse document frequency (TF–IDF) based on the bag-of-words scheme, topic distribution based on latent Dirichlet allocation (LDA), and neural-network-based document embedding known as document to vector (Doc2Vec). The experimental results demonstrate that the proposed MCT is robust to parameter changes and outperforms benchmark methods under various conditions.
| Original language | English |
|---|---|
| Pages (from-to) | 15-29 |
| Number of pages | 15 |
| Journal | Information Sciences |
| Volume | 477 |
| DOIs | |
| Publication status | Published - 2019 Mar 1 |
Fingerprint
Keywords
- Co-training
- Doc2vec
- Document classification
- LDA
- Semi-supervised learning
- TF–IDF
ASJC Scopus subject areas
- Software
- Control and Systems Engineering
- Theoretical Computer Science
- Computer Science Applications
- Information Systems and Management
- Artificial Intelligence
Cite this
Multi-co-training for document classification using various document representations : TF–IDF, LDA, and Doc2Vec. / Kim, Donghwa; Seo, Deokseong; Cho, Suhyoun; Kang, Pilsung.
In: Information Sciences, Vol. 477, 01.03.2019, p. 15-29.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - Multi-co-training for document classification using various document representations
T2 - TF–IDF, LDA, and Doc2Vec
AU - Kim, Donghwa
AU - Seo, Deokseong
AU - Cho, Suhyoun
AU - Kang, Pilsung
PY - 2019/3/1
Y1 - 2019/3/1
N2 - The purpose of document classification is to assign the most appropriate label to a specified document. The main challenges in document classification are insufficient label information and unstructured sparse format. A semi-supervised learning (SSL) approach could be an effective solution to the former problem, whereas the consideration of multiple document representation schemes can resolve the latter problem. Co-training is a popular SSL method that attempts to exploit various perspectives in terms of feature subsets for the same example. In this paper, we propose multi-co-training (MCT) for improving the performance of document classification. In order to increase the variety of feature sets for classification, we transform a document using three document representation methods: term frequency–inverse document frequency (TF–IDF) based on the bag-of-words scheme, topic distribution based on latent Dirichlet allocation (LDA), and neural-network-based document embedding known as document to vector (Doc2Vec). The experimental results demonstrate that the proposed MCT is robust to parameter changes and outperforms benchmark methods under various conditions.
AB - The purpose of document classification is to assign the most appropriate label to a specified document. The main challenges in document classification are insufficient label information and unstructured sparse format. A semi-supervised learning (SSL) approach could be an effective solution to the former problem, whereas the consideration of multiple document representation schemes can resolve the latter problem. Co-training is a popular SSL method that attempts to exploit various perspectives in terms of feature subsets for the same example. In this paper, we propose multi-co-training (MCT) for improving the performance of document classification. In order to increase the variety of feature sets for classification, we transform a document using three document representation methods: term frequency–inverse document frequency (TF–IDF) based on the bag-of-words scheme, topic distribution based on latent Dirichlet allocation (LDA), and neural-network-based document embedding known as document to vector (Doc2Vec). The experimental results demonstrate that the proposed MCT is robust to parameter changes and outperforms benchmark methods under various conditions.
KW - Co-training
KW - Doc2vec
KW - Document classification
KW - LDA
KW - Semi-supervised learning
KW - TF–IDF
UR - http://www.scopus.com/inward/record.url?scp=85055422604&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85055422604&partnerID=8YFLogxK
U2 - 10.1016/j.ins.2018.10.006
DO - 10.1016/j.ins.2018.10.006
M3 - Article
AN - SCOPUS:85055422604
VL - 477
SP - 15
EP - 29
JO - Information Sciences
JF - Information Sciences
SN - 0020-0255
ER -