Social Media Text Classification under Negative Covariate Shift

Abstract

In a typical social media content analysis task, the user is interested in analyzing posts of a particular topic. Identifying such posts is often formulated as a classification problem. However, this problem is challenging. One key issue is covariate shift. That is, the training data is not fully representative of the test data. We observed that the covariate shift mainly occurs in the negative data because topics discussed in social media are highly diverse and numerous, but the user-labeled negative training data may cover only a small number of topics. This paper proposes a novel technique to solve the problem. The key novelty of the technique is the transformation of document representation from the traditional ngram feature space to a center-based similarity (CBS) space. In the CBS space, the covariate shift problem is significantly mitigated, which enables us to build much better classifiers. Experiment results show that the proposed approach markedly improves classification.

Learning Cumulatively to Become More Knowledgeable

“…Second, it only focuses on improving accuracy on the classes with seed examples. Our work in [14] dealt with the problem using an entirely different approach adopted from [13]. However, these works did not propose or deal with cumulative learning, which is important for an intelligent system as it allows the system to learn more and more and become more and more knowledgeable.…”

“…At time + 1, the new dataset +1 of class +1 arrives, and the classification model needs to be updated or extended to produce a new classification model +1 . We note that each ℎ in or +1 is a 1-vs-rest SVM classifier built using the CBS learning method in [13] for class treating as the positive class. We will discuss CBS learning in the next section.…”

“…This section gives an overview of the CBS learning method given in [13]. In [14], is defined as the positively labeled area that is sufficiently far from the center of the positive training examples.…”

“…To build an open world classifier, a recent learning technique called center-based similarity space learning (CBS learning), which was originally proposed by Fei and Liu [13] for solving a negative covariate shift problem, is employed to give an initial solution to the proposed open space formulation, which significantly reduces the open space risk compared to that of [31].…”

Support vector machine parameter tuning based on particle swarm optimization metaheuristic

“…The authors in [44] reported that a linear SVM achieves the best results consistently to SVM with different kernels including SVM-Poly. The authors in [7,16,26] also reported the linear SVM efficiency for binary text classification. Unfortunately, manual hyperparameter selection still remains one of the practical application issues, while recent literature still does not provide any heuristic rules or rules of thumb for this task in [41].…”