David L. Waltz died on March 22, 2012 after suffering from brain cancer.

Dave was a good friend to Natural Language Engineering, and provided some sage advice when Roberto Garigliano and I started working on the proposed journal in the early 1990s; he subsequently agreed to serve as a founding editorial board member.

During last decade, machine learning and, in particular, statistical approaches have become more and more important for research in Natural Language Processing (NLP) and Computational Linguistics. Nowadays, most stakeholders of the field use machine learning, as it can significantly enhance both system design and performance. However, machine learning requires careful parameter tuning and feature engineering for representing language phenomena. The latter becomes more complex when the system input/output data is structured, since the designer has both to (i) engineer features for representing structure and model interdependent layers of information, which is usually a non-trivial task; and (ii) generate a structured output using classifiers, which, in their original form, were developed only for classification or regression. Research in empirical NLP has been tackling this problem by constructing output structures as a combination of the predictions of independent local classifiers, eventually applying post-processing heuristics to correct incompatible outputs by enforcing global properties. More recently, some advances of the statistical learning theory, namely structured output spaces and kernel methods, have brought techniques for directly encoding dependencies between data items in a learning algorithm that performs global optimization. Within this framework, this special issue aims at studying, comparing, and reconciling the typical domain/task-specific NLP approaches to structured data with the most advanced machine learning methods. In particular, the selected papers analyze the use of diverse structured input/output approaches, ranging from re-ranking to joint constraint-based global models, for diverse natural language tasks, i.e., document ranking, syntactic parsing, sequence supertagging, and relation extraction between terms and entities. Overall, the experience with this special issue shows that, although a definitive unifying theory for encoding and generating structured information in NLP applications is still far from being shaped, some interesting and effective best practice can be defined to guide practitioners in modeling their own natural language application on complex data.

An increasing number of researchers and practitioners in Natural Language Engineering face the prospect of having to work with entire texts, rather than individual sentences. While it is clear that text must have useful structure, its nature may be less clear, making it more difficult to exploit in applications. This survey of work on discourse structure thus provides a primer on the bases of which discourse is structured along with some of their formal properties. It then lays out the current state-of-the-art with respect to algorithms for recognizing these different structures, and how these algorithms are currently being used in Language Technology applications. After identifying resources that should prove useful in improving algorithm performance across a range of languages, we conclude by speculating on future discourse structure-enabled technology.

Natural Language Engineering (NLE) has enjoyed another year promoting research of applied Natural Language Processing and serving the research community in the field. We were particularly pleased to register an increasing number of submissions on a wide range of topics reflecting the growing importance of the field. We were also delighted to receive a number of submissions representing the new article types announced in 2010, such as surveys and squibs.

In source reordering the order of the source words is permuted to minimize word order differences with the target sentence and then fed to a translation model. Earlier work highlights the benefits of resolving long-distance reorderings as a pre-processing step to standard phrase-based models. However, the potential performance improvement of source reordering and its impact on the components of the subsequent translation model remain unexplored. In this paper we study both aspects of source reordering. We set up idealized source reordering (oracle) models with/without syntax and present our own syntax-driven model of source reordering. The latter is a statistical model of inversion transduction grammar (ITG)-like tree transductions manipulating a syntactic parse and working with novel conditional reordering parameters. Having set up the models, we report translation experiments showing significant improvement on three language pairs, and contribute an extensive analysis of the impact of source reordering (both oracle and model) on the translation model regarding the quality of its input, phrase-table, and output. Our experiments show that oracle source reordering has untapped potential in improving translation system output. Besides solving difficult reorderings, we find that source reordering creates more monotone parallel training data at the back-end, leading to significantly larger phrase tables with higher coverage of phrase types in unseen data. Unfortunately, this nice property does not carry over to tree-constrained source reordering. Our analysis shows that, from the string-level perspective, tree-constrained reordering might selectively permute word order, leading to larger phrase tables but without increase in phrase coverage in unseen data.

In dialogue systems it is important to label the dialogue turns with dialogue-related meaning. Each turn is usually divided into segments and these segments are labelled with dialogue acts (DAs). A DA is a representation of the functional role of the segment. Each segment is labelled with one DA, representing its role in the ongoing discourse. The sequence of DAs given a dialogue turn is used by the dialogue manager to understand the turn. Probabilistic models that perform DA labelling can be used on segmented or unsegmented turns. The last option is more likely for a practical dialogue system, but it provides poorer results. In that case, a hypothesis for the number of segments can be provided to improve the results. We propose some methods to estimate the probability of the number of segments based on the transcription of the turn. The new labelling model includes the estimation of the probability of the number of segments in the turn. We tested this new approach with two different dialogue corpora: SwitchBoard and Dihana. The results show that this inclusion significantly improves the labelling accuracy.

Authorship attribution methods aim to determine the author of a document, by using information gathered from a set of documents with known authors. One method of performing this task is to create profiles containing distinctive features known to be used by each author. In this paper, a new method of creating an author or document profile is presented that detects features considered distinctive, compared to normal language usage. This recentreing approach creates more accurate profiles than previous methods, as demonstrated empirically using a known corpus of authorship problems. This method, named recentred local profiles, determines authorship accurately using a simple ‘best matching author’ approach to classification, compared to other methods in the literature. The proposed method is shown to be more stable than related methods as parameter values change. Using a weighted voting scheme, recentred local profiles is shown to outperform other methods in authorship attribution, with an overall accuracy of 69.9% on the ad-hoc authorship attribution competition corpus, representing a significant improvement over related methods.

This paper introduces a novel ranking refinement approach based on relevance feedback for the task of document retrieval. We focus on the problem of ranking refinement since recent evaluation results from Information Retrieval (IR) systems indicate that current methods are effective retrieving most of the relevant documents for different sets of queries, but they have severe difficulties to generate a pertinent ranking of them. Motivated by these results, we propose a novel method to re-rank the list of documents returned by an IR system. The proposed method is based on a Markov Random Field (MRF) model that classifies the retrieved documents as relevant or irrelevant. The proposed MRF combines: (i) information provided by the base IR system, (ii) similarities among documents in the retrieved list, and (iii) relevance feedback information. Thus, the problem of ranking refinement is reduced to that of minimising an energy function that represents a trade-off between document relevance and inter-document similarity. Experiments were conducted using resources from four different tasks of the Cross Language Evaluation Forum (CLEF) forum as well as from one task of the Text Retrieval Conference (TREC) forum. The obtained results show the feasibility of the method for re-ranking documents in IR and also depict an improvement in mean average precision compared to a state of the art retrieval machine.

Automatic determination of synonyms and/or semantically related words has various applications in Natural Language Processing. Two mainstream paradigms to date, lexicon-based and distributional approaches, both exhibit pros and cons with regard to coverage, complexity, and quality. In this paper, we propose three novel methods—two rule-based methods and one machine learning approach—to identify synonyms from definition texts in a machine-readable dictionary. Extracted synonyms are evaluated in two extrinsic experiments and one intrinsic experiment. Evaluation results show that our pattern-based approach achieves best performance in one of the experiments and satisfactory results in the other, comparable to corpus-based state-of-the-art results.

A vast amount of usable electronic data is in the form of unstructured text. The relation extraction task aims to identify useful information in text (e.g. PersonW works for OrganisationX, GeneY encodes ProteinZ) and recode it in a format such as a relational database or RDF triplestore that can be more effectively used for querying and automated reasoning. A number of resources have been developed for training and evaluating automatic systems for relation extraction in different domains. However, comparative evaluation is impeded by the fact that these corpora use different markup formats and notions of what constitutes a relation. We describe the preparation of corpora for comparative evaluation of relation extraction across domains based on the publicly available ACE 2004, ACE 2005 and BioInfer data sets. We present a common document type using token standoff and including detailed linguistic markup, while maintaining all information in the original annotation. The subsequent reannotation process normalises the two data sets so that they comply with a notion of relation that is intuitive, simple and informed by the semantic web. For the ACE data, we describe an automatic process that automatically converts many relations involving nested, nominal entity mentions to relations involving non-nested, named or pronominal entity mentions. For example, the first entity is mapped from ‘one’ to ‘Amidu Berry’ in the membership relation described in ‘Amidu Berry, one half of PBS’. Moreover, we describe a comparably reannotated version of the BioInfer corpus that flattens nested relations, maps part-whole to part-part relations and maps n-ary to binary relations. Finally, we summarise experiments that compare approaches to generic relation extraction, a knowledge discovery task that uses minimally supervised techniques to achieve maximally portable extractors. These experiments illustrate the utility of the corpora.1

Usually unsupervised dependency parsers try to optimize the probability of a corpus by revising the dependency model that is assumed to have generated the corpus. In this paper we explore a different view in which a dependency structure is, among other things, a partial order on the nodes in terms of centrality or saliency. Under this assumption we directly model centrality and derive dependency trees from the ordering of words. The result is an approach to unsupervised dependency parsing that is very different from standard ones in that it requires no training data. The input words are ordered by centrality, and a parse is derived from the ranking using a simple deterministic parsing algorithm, relying on the universal dependency rules defined by Naseem et al. (Naseem, T., Chen, H., Barzilay, R., Johnson, M. 2010. Using universal linguistic knowledge to guide grammar induction. In Proceedings of Empirical Methods in Natural Language Processing, Boston, MA, USA, pp. 1234–44.). Our approach is evaluated on data from twelve different languages and is remarkably competitive.

This paper presents an investigation of part of speech (POS) tagging for Arabic as it occurs naturally, i.e. unvocalized text (without diacritics). We also do not assume any prior tokenization, although this was used previously as a basis for POS tagging. Arabic is a morphologically complex language, i.e. there is a high number of inflections per word; and the tagset is larger than the typical tagset for English. Both factors, the second one being partly dependent on the first, increase the number of word/tag combinations, for which the POS tagger needs to find estimates, and thus they contribute to data sparseness. We present a novel approach to Arabic POS tagging that does not require any pre-processing, such as segmentation or tokenization: whole word tagging. In this approach, the complete word is assigned a complex POS tag, which includes morphological information. A competing approach investigates the effect of segmentation and vocalization on POS tagging to alleviate data sparseness and ambiguity. In the segmentation-based approach, we first automatically segment words and then POS tags the segments. The complex tagset encompasses 993 POS tags, whereas the segment-based tagset encompasses only 139 tags. However, segments are also more ambiguous, thus there are more possible combinations of segment tags. In realistic situations, in which we have no information about segmentation or vocalization, whole word tagging reaches the highest accuracy of 94.74%. If gold standard segmentation or vocalization is available, including this information improves POS tagging accuracy. However, while our automatic segmentation and vocalization modules reach state-of-the-art performance, their performance is not reliable enough for POS tagging and actually impairs POS tagging performance. Finally, we investigate whether a reduction of the complex tagset to the Extra-Reduced Tagset as suggested by Habash and Rambow (Habash, N., and Rambow, O. 2005. Arabic tokenization, part-of-speech tagging and morphological disambiguation in one fell swoop. In Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL), Ann Arbor, MI, USA, pp. 573–80) will alleviate the data sparseness problem. While the POS tagging accuracy increases due to the smaller tagset, a closer look shows that using a complex tagset for POS tagging and then converting the resulting annotation to the smaller tagset results in a higher accuracy than tagging using the smaller tagset directly.

We present a general, novel methodology for extracting multi-word expressions (MWEs) of various types, along with their translations, from small, word-aligned parallel corpora. Unlike existing approaches, we focus on misalignments; these typically indicate expressions in the source language that are translated to the target in a non-compositional way. We introduce a simple algorithm that proposes MWE candidates based on such misalignments, relying on 1:1 alignments as anchors that delimit the search space. We use a large monolingual corpus to rank and filter these candidates. Evaluation of the quality of the extraction algorithm reveals significant improvements over naïve alignment-based methods. The extracted MWEs, with their translations, are used in the training of a statistical machine translation system, showing a small but significant improvement in its performance.

As described in this paper, we specifically examine the structural learning problem of a supertagging task. Supertagging is a task to assign the most probable lexical entry to each word in a sentence. A supertagger is extremely important for a lexicalized grammar parser because an accurate supertagger can greatly reduce lexical ambiguity in downstream parser. Supertagging is more challenging than conventional sequence labeling tasks (e.g., part-of-speech tagging). First, the supertags are numerous. Supertags are the lexical entries defined in a lexicalized grammar, which consists of rich syntactic/semantic information. Second, the inter-supertag relation is more complex. A proper supertag assignment is expected to be compatible with other supertag assignments in a sentence to construct a parse tree. Commonly used adjacent label features (e.g., first-order edge feature) in a sequence labeling model are too rough for the supertagging task. Long-range information is extremely important for the supertagging task. Two approaches to consider long-range information in a supertagger's training stage are proposed. Specifically, we propose a dependency-informed supertagger to use word-to-word dependency derived from a dependency parser and generate long-range features as soft constraints in the training. In the forest-guided supertagger, we constrain the classifier to learn in a grammar-satisfying space and use a CFG filter to impose grammar constraints for the update of model parameters. The experiments show that the proposed structure-guided supertaggers perform significantly better than the baseline supertaggers. Based on the improved supertaggers, the F-score of the final parser is also improved. Using the forest-guided supertagger in a shift-reduce HPSG parser, we achieved a competitive parsing performance of 89.31% F-score with higher parsing speed than that of a state-of-the-art HPSG parser.

In this article, we explore the task of mood classification for blog postings. We propose a novel approach that uses the hierarchy of possible moods to achieve better results than a standard machine learning approach. We also show that using sentiment orientation features improves the performance of classification. We used the Livejournal blog corpus as a data set to train and evaluate our method. We present extensive error analysis and discuss the difficulty of the task.

This paper presents a method for the composition of at-location with other semantic relations. The method is based on inference axioms that combine two semantic relations yielding another relation that otherwise is not expressed. An experimental study conducted on PropBank, WordNet, and eXtended WordNet shows that inferences have high accuracy. The method is applicable to combining other semantic relations and it is beneficial to many semantically intense applications.

Determining whether two terms have an ancestor relation (e.g. Toyota Camry and car) or a sibling relation (e.g. Toyota and Honda) is an essential component of textual inference in Natural Language Processing applications such as Question Answering, Summarization, and Textual Entailment. Significant work has been done on developing knowledge sources that could support these tasks, but these resources usually suffer from low coverage, noise, and are inflexible when dealing with ambiguous and general terms that may not appear in any stationary resource, making their use as general purpose background knowledge resources difficult. In this paper, rather than building a hierarchical structure of concepts and relations, we describe an algorithmic approach that, given two terms, determines the taxonomic relation between them using a machine learning-based approach that makes use of existing resources. Moreover, we develop a global constraint-based inference process that leverages an existing knowledge base to enforce relational constraints among terms and thus improves the classifier predictions. Our experimental evaluation shows that our approach significantly outperforms other systems built upon the existing well-known knowledge sources.

This article proposes clause-level evaluation detection, which is a fine-grained type of opinion mining, and describes an unsupervised lexicon building method for capturing domain-specific knowledge by leveraging the similar polarities of sentiments between adjacent clauses. The lexical entries to be acquired are called polar atoms, the minimum human-understandable syntactic structures that specify the polarity of clauses. As a hint to obtain candidate polar atoms, we use context coherency, the tendency for the same polarity to appear successively in a context. Using the overall density and precision of coherency in the corpus, the statistical estimation picks up appropriate polar atoms from among the candidates, without any manual tuning of the threshold values. The experimental results show that the precision of polarity assignment with the automatically acquired lexicon was 83 per cent on average, and our method is robust for corpora in diverse domains and for the size of the initial lexicon.

In this study, we investigate using unsupervised generative learning methods for subjectivity detection across different domains. We create an initial training set using simple lexicon information and then evaluate two iterative learning methods with a base naive Bayes classifier to learn from unannotated data. The first method is self-training, which adds instances with high confidence into the training set in each iteration. The second is a calibrated EM (expectation-maximization) method where we calibrate the posterior probabilities from EM such that the class distribution is similar to that in the real data. We evaluate both approaches on three different domains: movie data, news resource, and meeting dialogues, and we found that in some cases the unsupervised learning methods can achieve performance close to the fully supervised setup. We perform a thorough analysis to examine factors, such as self-labeling accuracy of the initial training set in unsupervised learning, the accuracy of the added examples in self-training, and the size of the initial training set in different methods. Our experiments and analysis show inherent differences across domains and impacting factors explaining the model behaviors.