This paper separates conditional parameter estimation, which consistently raises test set accuracy on statistical NLP tasks, from conditional model structures, such as the conditional Markov model used for maximum-entropy tagging, which tend to lower accuracy. Error analysis on part-of-speech tagging shows that the actual tagging errors made by the conditionally structured model derive not only from label bias, but also from other ways in which the independence assumptions of the conditional model structure are unsuited to linguistic sequences. The paper presents new word-sense disambiguation and POS tagging experiments, and integrates apparently conflicting reports from other recent work.

Common approaches to multi-label classification learn independent classifiers for each category, and employ ranking or thresholding schemes for classification. Because they do not exploit dependencies between labels, such techniques are only well-suited to problems in which categories are independent. However, in many domains labels are highly interdependent. This paper explores multilabel conditional random field (CRF) classification models that directly parameterize label co-occurrences in multi-label classification. Experiments show that the models outperform their singlelabel counterparts on standard text corpora. Even when multilabels are sparse, the models improve subset classification error by as much as 40%.

This paper investigates two elements of Maximum Entropy tagging: the use of a correction feature in the Generalised Iterative Scaling (GIS) estimation algorithm, and techniques for model smoothing. We show analytically and empirically that the correction feature, assumed to be required for the correctness of GIS, is unnecessary. We also explore the use of a Gaussian prior and a simple cutoff for smoothing. The experiments are performed with two tagsets: the standard Penn Treebank POS tagset and the larger set of lexical types from Combinatory Categorial Grammar. 1

A large and growing number of web pages display contextual advertising based on keywords automatically extracted from the text of the page, and this is a substantial source of revenue supporting the web today. Despite the importance of this area, little formal, published research exists. We describe a system that learns how to extract keywords from web pages for advertisement targeting. The system uses a number of features, such as term frequency of each

This paper reports the development of loglinear models for the disambiguation in wide-coverage HPSG parsing. The estimation of log-linear models requires high computational cost, especially with widecoverage grammars. Using techniques to reduce the estimation cost, we trained the models using 20 sections of Penn Treebank. A series of experiments empirically evaluated the estimation techniques, and also examined the performance of the disambiguation models on the parsing of real-world sentences.

We present a novel statistical approach to semantic parsing, WASP, for constructing a complete, formal meaning representation of a sentence. A semantic parser is learned given a set of sentences annotated with their correct meaning representations. The main innovation of WASP is its use of state-of-the-art statistical machine translation techniques. A word alignment model is used for lexical acquisition, and the parsing model itself can be seen as a syntax-based translation model. We show that WASP performs favorably in terms of both accuracy and coverage compared to existing learning methods requiring similar amount of supervision, and shows better robustness to variations in task complexity and word order.

Weighted finite-state transducers suffer from the lack of a training algorithm. Training is even harder for transducers that have been assembled via finite-state operations such as composition, minimization, union, concatenation, and closure, as this yields tricky parameter tying. We formulate a "parameterized FST" paradigm and give training algorithms for it, including a general bookkeeping trick ("expectation semirings") that cleanly and efficiently computes expectations and gradients.

We propose a novel method for using the World Wide Web to acquire trigram estimates for statistical language modeling. We submit an N-gram as a phrase query to web search engines. The search engines return the number of web pages containing the phrase, from which the N-gram count is estimated. The N-gram counts are then used to form web-based trigram probability estimates. We discuss the properties of such estimates, and methods to interpolate them with traditional corpus based trigram estimates. We show that the interpolated models improve speech recognition word error rate significantly over a small test set. 1.

The development of FrameNet, a large database of semantically annotated sentences, has primed research into statistical methods for semantic tagging. We advance previous work by adopting a Maximum Entropy approach and by using previous tag information to find the highest probability tag sequence for a given sentence. Further we examine the use of sentence level syntactic pattern features to increase performance. We analyze our strategy on both human annotated and automatically identified frame elements, and compare performance to previous work on identical test data. Experiments indicate a statistically significant improvement (p<0.01) of over 6%. 1

Abstract. In this paper we unify divergence minimization and statistical inference by means of convex duality. In the process of doing so, we prove that the dual of approximate maximum entropy estimation is maximum a posteriori estimation. Moreover, our treatment leads to stability and convergence bounds for many statistical learning problems. Finally, we show how an algorithm by Zhang can be used to solve this class of optimization problems efficiently. 1