While methods for comparing two learning algorithms on a single data set have been scrutinized for quite some time already, the issue of statistical tests for comparisons of more algorithms on multiple data sets, which is even more essential to typical machine learning studies, has been all but ignored. This article reviews the current practice and then theoretically and empirically examines several suitable tests. Based on that, we recommend a set of simple, yet safe and robust non-parametric tests for statistical comparisons of classifiers: the Wilcoxon signed ranks test for comparison of two classifiers and the Friedman test with the corresponding post-hoc tests for comparison of more classifiers over multiple data sets. Results of the latter can also be neatly presented with the newly introduced CD (critical difference) diagrams.

/pdf/statistical-comparisons-of-classifiers-over-multiple-data-18bm1e3eph.pdf

Statistical Comparisons of Classifiers over Multiple Data Sets

This paper presents an overview of the field of recommender systems and describes the current generation of recommendation methods that are usually classified into the following three main categories: content-based, collaborative, and hybrid recommendation approaches. This paper also describes various limitations of current recommendation methods and discusses possible extensions that can improve recommendation capabilities and make recommender systems applicable to an even broader range of applications. These extensions include, among others, an improvement of understanding of users and items, incorporation of the contextual information into the recommendation process, support for multicriteria ratings, and a provision of more flexible and less intrusive types of recommendations.

/pdf/toward-the-next-generation-of-recommender-systems-a-survey-4wwsv77rns.pdf

Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions

Statistical approaches to processing natural language text have become dominant in recent years This foundational text is the first comprehensive introduction to statistical natural language processing (NLP) to appear The book contains all the theory and algorithms needed for building NLP tools It provides broad but rigorous coverage of mathematical and linguistic foundations, as well as detailed discussion of statistical methods, allowing students and researchers to construct their own implementations The book covers collocation finding, word sense disambiguation, probabilistic parsing, information retrieval, and other applications

/pdf/foundations-of-statistical-natural-language-processing-1ou5e6c2q6.pdf

Foundations of Statistical Natural Language Processing

This paper explores the use of Support Vector Machines (SVMs) for learning text classifiers from examples. It analyzes the particular properties of learning with text data and identifies why SVMs are appropriate for this task. Empirical results support the theoretical findings. SVMs achieve substantial improvements over the currently best performing methods and behave robustly over a variety of different learning tasks. Furthermore they are fully automatic, eliminating the need for manual parameter tuning.

/pdf/text-categorization-with-suport-vector-machines-learning-1onl65lfqq.pdf

Text Categorization with Suport Vector Machines: Learning with Many Relevant Features

The automated categorization (or classification) of texts into predefined categories has witnessed a booming interest in the last 10 years, due to the increased availability of documents in digital form and the ensuing need to organize them. In the research community the dominant approach to this problem is based on machine learning techniques: a general inductive process automatically builds a classifier by learning, from a set of preclassified documents, the characteristics of the categories. The advantages of this approach over the knowledge engineering approach (consisting in the manual definition of a classifier by domain experts) are a very good effectiveness, considerable savings in terms of expert labor power, and straightforward portability to different domains. This survey discusses the main approaches to text categorization that fall within the machine learning paradigm. We will discuss in detail issues pertaining to three different problems, namely, document representation, classifier construction, and classifier evaluation.

/pdf/machine-learning-in-automated-text-categorization-e4pkjhyj3r.pdf

Machine learning in automated text categorization

In this paper, we compare learning techniques based on statistical classification to traditional methods of relevance feedback for the document routing problem. We consider three classification techniques which have decision rules that are derived via explicit error minimization: linear discriminant analysis, logistic regression, and neural networks. We demonstrate that the classifiers perform 1015% better than relevance feedback via Rocchio expansion for the TREC-2 and TREC-3 routing tasks. Error minimization is difficult in high-dimensional feature spaces because the convergence process is slow and the models are prone to overfitting. We use two different strategies, latent semantic indexing and optimal term selection, to reduce the number of features. Our results indicate that features based on latent semantic indexing are more effective for techniques such as linear discriminant analysis and logistic regression, which have no way to protect against overfitting. Neural networks perform equally well with either set of features and can take advantage of the additional information available when both feature sets are used as input.

A comparison of classifiers and document representations for the routing problem

The majority of information retrieval experiments are evaluated by measures such as average precision and average recall. Fundamental decisions about the superiority of one retrieval technique over another are made solely on the basis of these measures. We claim that average performance figures need to be validated with a careful statistical analysis and that there is a great deal of additional information that can be uncovered by looking closely at the results of individual queries. This article is a case study of stemming algorithms which describes a number of novel approaches to evaluation and demonstrates their value. © 1996 John Wiley & Sons, Inc.

Stemming algorithms: a case study for detailed evaluation

The multilingual information retrieval system of the future will need to be able to retrieve documents across language boundaries. This extension of the classical IR problem is particularly challenging, ax significant resources are required to perform query translation. At Xerox, we are working to build a multilingual IR system and conducting a series of experiments to understand what factors are most important in making the system work. Using translated queries and a bilingual transfer dictionary, we have learned that crosslartguage multilingual IR is feasible, although performance lags considerably behind the monolingual standard. The experiments suggest that correct identification and translation of multi-word terminology is the single most important source of error in the system, although amblguit y in translation also contributes to poor performance.

Querying across languages: a dictionary-based approach to multilingual information retrieval

The TREC–9 filtering track measures the ability of systems to build persistent user profiles which successfully separate relevant and non-relevant documents. It consists of three major subtasks: adaptive filtering, batch filtering, and routing. In adaptive filtering, the system begins with only a topic statement and a small number of positive examples, and must learn a better profile from on-line feedback. Batch filtering and routing are more traditional machine learning tasks where the system begins with a large sample of evaluated training documents. This report describes the track, presents some evaluation results, and provides a general commentary on lessons learned from this year’s track.

/pdf/the-trec-9-filtering-track-final-report-bwxyb7qour.pdf

The TREC-9 Filtering Track Final Report.

There is strong empirical and theoretic evidence that combination of retrieval methods can improve performance. In this paper, we systematically compare combination strategies in the context of document filtering, using queries from the Tipster reference corpus. We find that simple averaging strategies do indeed improve performance, but that direet averaging of probability estimates is not the correet approach. Instead, the probabiJit y estimates must be renormalized using logistic regression on the known relevance judgments. We examine more complex combination strat~ gies but find them less successful due to the high correlations among our filtering methods which are optimized over the same training data and employ similar document represerttations.

David A. Hull

Papers

A comparison of classifiers and document representations for the routing problem

Stemming algorithms: a case study for detailed evaluation

Querying across languages: a dictionary-based approach to multilingual information retrieval

The TREC-9 Filtering Track Final Report.

Method combination for document filtering