Abstract. In this paper we present methods of enhancing existing discriminative classifiers for multi-labeled predictions. Discriminative methods like support vector machines perform very well for uni-labeled text classification tasks. Multi-labeled classification is a harder task subject to relatively less attention. In the multi-labeled setting, classes are often related to each other or part of a is-a hierarchy. We present a new technique for combining text features and features indicating relationships between classes, which can be used with any discriminative algorithm. We also present two enhancements to the margin of SVMs for building better models in the presence of overlapping classes. We present results of experiments on real world text benchmark datasets. Our new methods beat accuracy of existing methods with statistically significant improvements.

Noise is a stark reality in real life data. Especially in the domain of text analytics, it has a significant impact as data cleaning forms a very large part of the data processing cycle. Noisy unstructured text is common in informal settings such as on-line chat, SMS, email, newsgroups and blogs, automatically transcribed text from speech, and automatically recognized text from printed or handwritten material. Gigabytes of such data is being generated everyday on the Internet, in contact centers, and on mobile phones. Researchers have looked at various text mining issues such as pre-processing and cleaning noisy text, information extraction, rule learning, and classification for noisy text. This paper focuses on the issues faced by automatic text classifiers in analyzing noisy documents coming from various sources. The goal of this paper is to bring out and study the effect of different kinds of noise on automatic text classification. Does the nature of such text warrant moving beyond traditional text classification techniques? We present detailed experimental results with simulated noise on the Reuters-21578 and 20-newsgroups benchmark datasets. We present interesting results on real-life noisy datasets from various CRM domains.

Support vector machines (SVMs) excel at two-class discriminative learning problems. They often outperform generative classifiers, especially those that use inaccurate generative models, such as the naive Bayes (NB) classifier. On the other hand, generative classifiers have no trouble in handling an arbitrary number of classes efficiently, and NB classifiers train much faster than SVMs owing to their extreme simplicity. In contrast, SVMs handle multi-class problems by learning redundant yes/no (one-vs-others) classifiers for each class, further worsening the performance gap. We propose a new technique for multi-way classification which exploits the accuracy of SVMs and the speed of NB classifiers. We first use a NB classifier to quickly compute a confusion matrix, which is used to reduce the number and complexity of the two-class SVMs that are built in the second stage. During testing, we first get the prediction of a NB classifier and use that to selectively apply only a subset of the two-class SVMs. On standard benchmarks, our algorithm is 3 to 6 times faster than SVMs and yet matches or even exceeds their accuracy.

Effective incorporation of human expertise, while exerting a low cognitive load, is a critical aspect of real-life text classification applications that is not adequately addressed by batch-supervised highaccuracy learners. Standard text classifiers are supervised in only one way: assigning labels to whole documents. They are thus deprived of the enormous wisdom that humans carry about the significance of words and phrases in context. We present HIClass, an interactive and exploratory labeling package that actively collects user opinion on feature representations and choices, as well as whole-document labels, while minimizing redundancy in the input sought. Preliminary experience suggests that, starting with essentially an unlabeled corpus, very little cognitive labor suffices to set up a labeled collection on which standard classifiers perform well.

Abstract. Semi-supervised clustering models, that incorporate user provided constraints to yield meaningful clusters, have recently become a popular area of research. In this paper, we propose a cluster-level semi-supervision model for inter-active clustering. Prototype based clustering algorithms typically alternate between updating cluster descriptions and assignment of data items to clusters. In our model, the user provides semi-supervision directly for these two steps. Assignment feedback re-assigns data items among existing clusters, while cluster description feedback helps to position existing cluster centers more meaningfully. We argue that providing such supervision is more natural for exploratory data mining, where the user discovers and interprets clusters as the algorithm progresses, in comparison to the pair-wise instance level supervision model, particularly for high dimensional data such as document collection. We show how such feedback can be interpreted as constraints and incorporated within the kmeans clustering framework. Using experimental results on multiple real-world datasets, we show that this framework improves clustering performance significantly beyond traditional k-means. Interestingly, when given the same number of feedbacks from the user, the proposed framework significantly outperforms the pair-wise supervision model.