Applying information extraction techniques in the biological domain has been a growing research area over the past few years. Numerous large scale copora have been developed [10] or are being developed [4]

We present a novel method for discovering parallel sentences in comparable, non-parallel corpora. We train a maximum entropy classifier that, given a pair of sentences, can reliably determine whether or not they are translations of each other. Using this approach, we extract parallel data from large Chinese, Arabic, and English non-parallel newspaper corpora. We evaluate the quality of the extracted data by showing that it improves the performance of a state-of-the-art statistical machine translation system. We also show that a good-quality MT system can be built from scratch by starting with a very small parallel corpus (100,000 words) and exploiting a large non-parallel corpus. Thus, our method can be applied with great benefit to language pairs for which only scarce resources are available. 1.

In this paper we present the Alpino Dependency Treebank and the tools that we have developed to facilitate the annotation process. Annotation typically starts with parsing a sentence with the Alpino parser, a wide coverage parser of Dutch text. The number of parses that is generated is reduced through interactive lexical analysis and constituent marking. A tool for on line addition of lexical information facilitates the parsing of sentences with unknown words. The selection of the best parse is done efficiently with the parse selection tool. At this moment, the Alpino Dependency Treebank consists of about 6,000 sentences of newspaper text that are annotated with dependency trees. The corpus can be used for linguistic exploration as well as for training and evaluation purposes.

Abstract. In this paper, we propose the use of the Maximum Entropy approach for the task of automatic image annotation. Given labeled training data, Maximum Entropy is a statistical technique which allows one to predict the probability of a label given test data. The techniques allow for relationships between features to be effectively captured. and has been successfully applied to a number of language tasks including machine translation. In our case, we view the image annotation task as one where a training data set of images labeled with keywords is provided and we need to automatically label the test images with keywords. To do this, we first represent the image using a language of visterms and then predict the probability of seeing an English word given the set of visterms forming the image. Maximum Entropy allows us to compute the probability and in addition allows for the relationships between visterms to be incorporated. The experimental results show that Maximum Entropy outperforms one of the classical translation models that has been applied to this task and the Cross Media Relevance Model. Since the Maximum Entropy model allows for the use of a large number of predicates to possibly increase performance even further, Maximum Entropy model is a promising model for the task of automatic image annotation. 1

This paper describes discriminative language modeling for a large vocabulary speech recognition task. We contrast two parameter estimation methods: the perceptron algorithm, and a method based on conditional random fields (CRFs). The models are encoded as deterministic weighted finite state automata, and are applied by intersecting the automata with word-lattices that are the output from a baseline recognizer. The perceptron algorithm has the benefit of automatically selecting a relatively small feature set in just a couple of passes over the training data. However, using the feature set output from the perceptron algorithm (initialized with their weights), CRF training provides an additional 0.5 % reduction in word error rate, for a total 1.8 % absolute reduction from the baseline of 39.2%. 1

Abstract. We consider the problem of estimating an unknown probability distribution from samples using the principle of maximum entropy (maxent). To alleviate overfitting with a very large number of features, we propose applying the maxent principle with relaxed constraints on the expectations of the features. By convex duality, this turns out to be equivalent to finding the Gibbs distribution minimizing a regularized version of the empirical log loss. We prove nonasymptotic bounds showing that, with respect to the true underlying distribution, this relaxed version of maxent produces density estimates that are almost as good as the best possible. These bounds are in terms of the deviation of the feature empirical averages relative to their true expectations, a number that can be bounded using standard uniform-convergence techniques. In particular, this leads to bounds that drop quickly with the number of samples, and that depend very moderately on the number or complexity of the features. We also derive and prove convergence for both sequential-update and parallel-update algorithms. Finally, we briefly describe experiments on data relevant to the modeling of species geographical distributions. 1

Large-scale logistic regression arises in many applications such as document classification and natural language processing. In this paper, we apply a trust region Newton method to maximize the log-likelihood of the logistic regression model. The proposed method uses only approximate Newton steps in the beginning, but achieves fast convergence in the end. Experiments show that it is faster than the commonly used quasi Newton approach for logistic regression. We also compare it with existing linear SVM implementations. 1

The study of phonotactics (e.g., the ability of English speakers to distinguish possible words like blick from impossible words like *bnick) is a central topic in phonology. We propose a theory of phonotactic grammars and a learning algorithm that constructs such grammars from positive evidence. Our grammars consist of constraints that are assigned numerical weights according to the principle of maximum entropy. Possible words are assessed by these grammars based on the weighted sum of their constraint violations. The learning algorithm yields grammars that can capture both categorical and gradient phonotactic patterns. The algorithm is not provided with any constraints in advance, but uses its own resources to form constraints and weight them. A baseline model, in which Universal Grammar is reduced to a feature set and an SPE-style constraint format, suffices to learn many phonotactic phenomena. In order to learn nonlocal phenomena such as stress and vowel harmony, it is necessary to augment the model with autosegmental tiers and metrical grids. Our results thus offer novel, learning-theoretic support for such representations. We apply the model to English syllable onsets, Shona vowel harmony, quantity-insensitive stress typology, and the full phonotactics of Wargamay, showing that the learned grammars capture the distributional generalizations of these languages and accurately predict the findings of a phonotactic experiment.

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.

Dimensionality reduction of empirical co-occurrence data is a fundamental problem in unsupervised learning. It is also a well studied problem in statistics known as the analysis of cross-classified data. One principled approach to this problem is to represent the data in low dimension with minimal loss of (mutual) information contained in the original data. In this paper we introduce an information theoretic nonlinear method for finding such a most informative dimension reduction. In contrast with...