We demonstrate the effectiveness of multilingual learning for unsupervised part-of-speech tagging. The central assumption of our work is that by combining cues from multiple languages, the structure of each becomes more apparent. We consider two ways of applying this intuition to the problem of unsupervised part-of-speech tagging: a model that directly merges tag structures for a pair of languages into a single sequence and a second model which instead incorporates multilingual context using latent variables. Both approaches are formulated as hierarchical Bayesian models, using Markov Chain Monte Carlo sampling techniques for inference. Our results demonstrate that by incorporating multilingual evidence we can achieve impressive performance gains across a range of scenarios. We also found that performance improves steadily as the number of available languages increases.

We investigate the global GRAMMAR constraint over restricted classes of context free grammars like deterministic and unambiguous context-free grammars. We show that detecting disentailment for the GRAMMAR constraint in these cases is as hard as parsing an unrestricted context free grammar.We also consider the class of linear grammars and give a propagator that runs in quadratic time. Finally, to demonstrate the use of linear grammars, we show that a weighted linear GRAMMAR constraint can efficiently encode the EDITDISTANCE constraint, and a conjunction of the EDITDISTANCE constraint and the REGULAR constraint

While much work on learning in planning focused on learning domain physics (i.e., action models), and search control knowledge, little attention has been paid towards learning user preferences on desirable plans. Hierarchical task networks (HTN) are known to provide an effective way to encode user prescriptions about what constitute good plans. However, manual construction of these methods is complex and error prone. In this paper, we propose a novel approach to learning probabilistic hierarchical task networks that capture user preferences by examining user-produced plans given no prior information about the methods (in contrast, most prior work on learning within the HTN framework focused on learning “method preconditions”—i.e., domain physics—assuming that the structure of the methods is given as input). We will show that this problem has close parallels to the problem of probabilistic grammar induction, and describe how grammar induction methods can be adapted to learn task networks. We will empirically demonstrate the effectiveness of our approach by showing that task networks we learn are able to generate plans with a distribution close to the distribution of the userpreferred plans.

This paper investigates a generative history-based parsing model that synchronises the derivation of non-planar graphs representing semantic dependencies with the derivation of dependency trees representing syntactic structures. To process non-planarity online, the semantic transition-based parser uses a new technique to dynamically reorder nodes during the derivation. While the synchronised derivations allow different structures to be built for the semantic non-planar graphs and syntactic dependency trees, useful statistical dependencies between these structures are modeled using latent variables. The resulting synchronous parser achieves competitive performance on the CoNLL-2008 shared task, achieving relative error reduction of 12% in semantic F score over previously proposed synchronous models that cannot process non-planarity online.

Web-scale data has been used in a diverse range of language research. Most of this research has used web counts for only short, fixed spans of context. We present a unified view of using web counts for lexical disambiguation. Unlike previous approaches, our supervised and unsupervised systems combine information from multiple and overlapping segments of context. On the tasks of preposition selection and context-sensitive spelling correction, the supervised system reduces disambiguation error by 20-24% over the current state-of-the-art.

Arabic Subject Markers are interface phenomena (specifically between morphology and syntax). In this paper, I describe them briefly, I give my linguistic analysis within the framework of the Head-Driven Phrase Structure Grammar and I show how I implement them in the LKB system. I show that this system, despite its strength, does not allow for a proper implementation of these units.

We present an approach to patent claim simplification which segments claim sentences into clausal discourse units, transforms them into complete sentences, establishes coreference relations and builds a discourse structure between discourse units. The four stages are necessary to allow for the syntactic analysis of otherwise unparsable claim sentences and their regeneration using discourse structure and coreference relations in order to ensure the production of a cohesive and coherent paraphrase/summary.

The education community is moving towards constructed or free-text responses and computer-based assessment. At the same time, progress in natural language processing and knowledge representation has made it possible to consider free-text or constructed responses without having to fully understand the text. c-rater is a technology at Educational Testing Service (ETS) used for automatic content scoring for short, free-text responses. This paper describes some of the major developments made in c-rater recently.

This study evaluates the robustness of established computational indices used to assess text relatedness in user-language. The original User-Language Paraphrase Corpus (ULPC) was compared to a corrected version, in which each paraphrase was corrected for typographical and grammatical errors. Error correction significantly affected values for each of five computational indices, indicating greater similarity of the target sentence to the corrected paraphrase than to the original paraphrase. Moreover, misspelled target words accounted for a large proportion of the differences. This study also evaluated potential effects on correlations between computational indices and human ratings of paraphrases. The corrections did not yield assessments that were any more or less comparable to trained human raters than were the original paraphrases containing typographical or grammatical errors. The results suggest that although correcting for errors may optimize certain computational indices, the corrections are not necessary for comparing the indices to expert ratings.

A discriminative large margin classifier based approach to anaphora resolution for neuroscience abstracts is presented. The system employs both syntactic and semantic features. A support vector machine based word sense disambiguation method combining evidence from three methods, that use WordNet and Wikipedia, is also introduced and used for semantic features. The support vector machine anaphora resolution classifier with probabilistic outputs achieved almost four-fold improvement in accuracy over the baseline method.