Self-Taught convolutional neural networks for short text clustering.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Natural Language Processing

Background:

• Short text clustering presents challenges due to sparse data representations.
• Existing methods often struggle to incorporate rich semantic features effectively.
• Unsupervised learning of deep text representations remains an active research area.

Purpose of the Study:

• To propose a flexible Self-Taught Convolutional neural network framework for Short Text Clustering (STC²).
• To enhance the incorporation of semantic features and learn unbiased deep text representations in an unsupervised manner.
• To improve the accuracy and flexibility of short text clustering.

Main Methods:

• Embedding raw text features into compact binary codes using unsupervised dimensionality reduction.
• Utilizing word embeddings with convolutional neural networks to learn deep feature representations.
• Employing K-means clustering on the learned representations for final cluster identification.

Main Results:

• The proposed STC² framework successfully incorporates semantic features and learns deep representations.
• Experimental results show the framework's effectiveness and flexibility.
• The method outperforms several popular clustering techniques on public datasets.

Conclusions:

• The Self-Taught Convolutional neural network framework (STC²) offers a robust solution for short text clustering.
• The approach effectively addresses the sparsity issue in short text representation.
• STC² demonstrates superior performance compared to existing unsupervised clustering methods.