Self-Taught Convolutional Neural Networks for Short Text Clustering

@article{xu2017self,
title={Self-Taught Convolutional Neural Networks for Short Text Clustering},
author={Xu, Jiaming and Xu, Bo and Wang, Peng and Zheng, Suncong and Tian, Guanhua and Zhao, Jun and Xu, Bo},
journal={Neural Networks},
year={2017}
}

Note that:

Here are instructions of the demo dataset&software for the paper [Self-Taught Convolutional Neural Networks for Short Text Clustering]

Usage:

Please download the software and dataset packages, and put them into one folder;

The main function: ./software/main_STC2.m, please first "cd ./software/" and then run main_STC2.m via matlab;

Notices:

The suggested memory of machine is 16GB RAM;

The suggested matlab version is R2011 and above;

This is a demo package which includes the all details about porposed method and baselines;

K-means clustering is very slow on original high-dimensionality (2W~3W dim.) text features;
If you want to run clustering via Kmeans, please have a little patience, and we strongly suggest that you directly refer the KMeans results in our paper which reports the average results by running KMeans 500 times;

Please feel free to send me emails if you have any problems in using this package.

Instructions of Archives:

./README.md: Some notices and instructions.
./dataset/

-- Biomedical.txt: the raw 20,000 short text;
-- Biomedical_gnd.txt: the labels;
-- Biomedical_vocab2idx.dic: vocabulary index;
-- Biomedical_index.txt: has transfered the words into idx;
-- Biomedical-lite.mat: mini dataset only including feature vectors (fea) and labels (gnd);
-- Biomedical-STC2.mat: dataset for STC^2, including 20,000 short texts, 20 topics/tags and the pre-trained word embeddings;
-- SearchSnippets.txt: the raw 12,340 short text;
-- SearchSnippets_vocab2idx.dic: vocabulary index;
-- SearchSnippets_index.txt: has transfered the words into idx;
-- SearchSnippets-lite.mat: mini dataset only including feature vectors (fea) and labels (gnd);
-- SearchSnippets-STC2.mat: dataset for STC^2, including 12,340 short texts, 8 topics/tags and the pre-trained word embeddings;
-- StackOverflow.txt: the raw 20,000 short text;
-- StackOverflow_gnd.txt: the labels;
-- StackOverflow_vocab2idx.dic: vocabulary index;
-- StackOverflow_index.txt: has transfered the words into idx;
-- StackOverflow-lite.mat: mini dataset only including feature vectors (fea) and labels (gnd);
-- StackOverflow-STC2.mat: dataset for STC^2, including 20,000 short texts, 20 topics/tags and the pre-trained word embeddings;

./software/: Main folder of software;

-- main_STC2.m: main function, and select one clustering method here: Kmeans, RecNN, AveEmbedding, LSA, Spectral_LE, etc.;
-- run.sh: running it on commond line for linux user rather than window user;
-- STC2.m: interfaces of clustering methods;
-- STC2_CNN.m: interfaces of DCNN;
-- AE/: Average Embedding (AE) folder;
-- DCNN/: Dynamic Convolutional Neural Network (DCNN)[1] folder;
-- LE/: Laplacian Eigenmaps (LE)[2] folder;
-- LPI/: Locality Preserving Indexing (LPI)[3] folder;
-- LSA/: Latent Semantic Analysis (LSA)[4] folder;
-- Para2vec/: Paragraph vector (Para2vec)[5] folder;
-- RecNN/: Recursive Neural Network (RecNN)[6] folder;
-- results/: All evaluate results (ACC and NMI) of clustering will be saved in this folder;
-- tools/: Tool folder;
-- benchmarks/: Contains some classification benchmarks, SVM-linear or SVM-RBF on TF, TFIDF or AE. Get more classification details into this folder.

References:

[1]. N. Kalchbrenner, E. Grefenstette, P. Blunsom, A convolutional neural network for modelling sentences, ACL, 2014.
[2]. M. Belkin, P. Niyogi, Laplacian eigenmaps and spectral techniques for embedding and clustering, NIPS, 2001.
[3]. D. Cai, X. He, J. Han, Document clustering using locality preserving indexing, IEEE Transactions on Knowledge and Data Engineering, 2005.
[4]. S. C. Deerwester, S. T. Dumais, T. K. Landauer, G. W. Furnas, R. A. Harshman, Indexing by latent semantic analysis, JAsIs, 1990.
[5]. Q. Le, T. Mikolov, Distributed representations of sentences and documents, ICML, 2014.
[6]. R. Socher, J. Pennington, E. H. Huang, A. Y. Ng, C. D. Manning, Semisupervised recursive autoencoders for predicting sentiment distributions, EMNLP, 2011.

mbelousov/STC2

Self-Taught Convolutional Neural Networks for Short Text Clustering

Note that: