Yinuo Wang, Baiyang Wang, Robert Mercer, Frank Rudzicz, Sudipta Singha Roy, Pengjie Ren, Zhumin Chen, Xindi Wang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Scientific paper summarization has been the focus of much recent research. Unlike previous research which summarizes only the paper in question, or which summarizes the paper and the papers that it references, or which summarizes the paper and the citing sentences from the papers that cite it, this work puts all three of these summarization techniques together. To accomplish this, we have, by utilizing the citation network, introduced a corpus for scientific document summarization that provides information about the document being summarized, the papers referenced by it, as well as the papers that have cited it. The proposed summarizer model utilizes the referenced articles as background information and citing articles to capture the impact of the scientific document on the research community. Another aspect of the proposed model is its ability to generate both the extractive and abstractive summaries in parallel. The parallel training helps the counterparts to improve their individual performance. Results have shown that the summaries are of high quality when considering the standard metrics.

The International Classification of Diseases (ICD) is an authoritative medical classification system of different diseases and conditions for clinical and management purposes. ICD indexing aims to assign a subset of ICD codes to a medical record. Since human coding is labour-intensive and error-prone, many studies employ machine learning techniques to automate the coding process. ICD coding is a challenging task, as it needs to assign multiple codes to each medical document from an extremely large hierarchically organized collection. In this paper, we propose a novel approach for ICD indexing that adopts three ideas: (1) we use a multi-level deep dilated residual convolution encoder to aggregate the information from the clinical notes and learn document representations across different lengths of the texts; (2) we formalize the task of ICD classification with auxiliary knowledge of the medical records, which incorporates not only the clinical texts but also different clinical code terminologies and drug prescriptions for better inferring the ICD codes; and (3) we introduce a graph convolutional network to leverage the co-occurrence patterns among ICD codes, aiming to enhance the quality of label representations. Experimental results show the proposed method achieves state-of-the-art performance on a number of measures.

Anemily Machina, Robert Mercer. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Xindi Wang, Robert Mercer, Frank Rudzicz. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Large language models (LLMs) have shown remarkable capacity for in-context learning (ICL), where learning a new task from just a few training examples is done without being explicitly pre-trained. However, despite the success of LLMs, there has been little understanding of how ICL learns the knowledge from the given prompts. In this paper, to make progress toward understanding the learning behaviour of ICL, we train the same LLMs with the same demonstration examples via ICL and supervised learning (SL), respectively, and investigate their performance under label perturbations (i.e., noisy labels and label imbalance) on a range of classification tasks. First, via extensive experiments, we find that gold labels have significant impacts on the downstream in-context performance, especially for large language models; however, imbalanced labels matter little to ICL across all model sizes. Second, when comparing with SL, we show empirically that ICL is less sensitive to label perturbations than SL, and ICL gradually attains comparable performance to SL as the model size increases.

Currently, Medical Subject Headings (MeSH) are manually assigned to every biomedical article published and subsequently recorded in the PubMed database to facilitate retrieving relevant information.With the rapid growth of the PubMed database, large-scale biomedical document indexing becomes increasingly important.MeSH indexing is a challenging task for machine learning, as it needs to assign multiple labels to each article from an extremely large hierachically organized collection.To address this challenge, we propose KenMeSH, an end-to-end model that combines new text features and a dynamic Knowledge-enhanced mask attention that integrates document features with MeSH label hierarchy and journal correlation features to index MeSH terms.Experimental results show the proposed method achieves state-of-the-art performance on a number of measures.

In the field of Natural Language Processing (NLP), extracting method entities from biomedical text has been a challenging task. Scientific research papers commonly consist of complex keywords and domain-specific terminologies, and new terminologies are continuously appearing. In this research, we find method terminologies in biomedical text using both rule-based and machine learning techniques. We first use linguistic features to extract method sentence candidates from a large corpus of biomedical text. Then, we construct a silver standard biomedical corpus composed of these sentences. With a rule-based method that makes use of the Stanza dependency parsing module, we label the method entities in these sentences. Using this silver standard corpus we train two machine learning algorithms to automatically extract method entities from biomedical text. Our results show that it is possible to develop machine learning models that can automatically extract method entities to a reasonable accuracy without the need for a gold standard dataset.

Due to the success of pre-trained language models, versions of languages other than English have been released in recent years. This fact implies the need for resources to evaluate these models. In the case of Spanish, there are few ways to systematically assess the models’ quality. In this paper, we narrow the gap by building two evaluation benchmarks. Inspired by previous work (Conneau and Kiela, 2018; Chen et al., 2019), we introduce Spanish SentEval and Spanish DiscoEval, aiming to assess the capabilities of stand-alone and discourse-aware sentence representations, respectively. Our benchmarks include considerable pre-existing and newly constructed datasets that address different tasks from various domains. In addition, we evaluate and analyze the most recent pre-trained Spanish language models to exhibit their capabilities and limitations. As an example, we discover that for the case of discourse evaluation tasks, mBERT, a language model trained on multiple languages, usually provides a richer latent representation than models trained only with documents in Spanish. We hope our contribution will motivate a fairer, more comparable, and less cumbersome way to evaluate future Spanish language models.

Citations are frequently used in publications to support the presented results and to demonstrate the previous discoveries while also assisting the reader in following the chronological progression of information through publications. In scientific publications, a citation refers to the referenced document, but it makes no mention of the exact span of text that is being referred to. Connecting the citation to this span of text is called citation linkage. In this paper, to find these citation linkages in biomedical research publications using deep learning, we provide a synthetic silver standard corpus as well as the method to build this corpus. The motivation for building this corpus is to provide a training set for deep learning models that will locate the text spans in a reference article, given a citing statement, based on semantic similarity. This corpus is composed of sentence pairs, where one sentence in each pair is the citing statement and the other one is a candidate cited statement from the referenced paper. The corpus is annotated using an unsupervised sentence embedding method. The effectiveness of this silver standard corpus for training citation linkage models is validated against a human-annotated gold standard corpus.

Medical Subject Heading (MeSH) indexing refers to the problem of assigning a given biomedical document with the most relevant labels from an extremely large set of MeSH terms. Currently, the vast number of biomedical articles in the PubMed database are manually annotated by human curators, which is time consuming and costly; therefore, a computational system that can assist the indexing is highly valuable. When developing supervised MeSH indexing systems, the availability of a large-scale annotated text corpus is desirable. A publicly available, large corpus that permits robust evaluation and comparison of various systems is important to the research community. We release a large scale annotated MeSH indexing corpus, MeSHup, which contains 1,342,667 full text articles, together with the associated MeSH labels and metadata, authors and publication venues that are collected from the MEDLINE database. We train an end-to-end model that combines features from documents and their associated labels on our corpus and report the new baseline.

The emergence of social networking sites has paved the way for researchers to collect and analyze massive data volumes. Twitter, being one of the leading micro-blogging sites worldwide, provides an excellent opportunity for its users to express their states of mind via short text messages known as tweets. Much research focusing on identifying emotions and sentiments conveyed through tweets has been done. We propose a BERT model fine-tuned to the emotion recognition and sentiment classification tasks and show that it performs better than previous models on standard datasets. We also explore the effectiveness of data augmentation and data enrichment for these tasks.

In natural language processing, the performance of a semantic similarity task relies heavily on the availability of a large corpus. Various monolingual corpora are available (mainly English); but multilingual resources are very limited. In this work, we describe a semi-automated framework to create a multilingual corpus which can be used for the multilingual semantic similarity task. The similar sentence pairs are obtained by crawling bilingual websites, whereas the dissimilar sentence pairs are selected by applying topic modeling and an Open-AI GPT model on the similar sentence pairs. We focus on websites in the government, insurance, and banking domains to collect English-French and English-Spanish sentence pairs; however, this corpus creation approach can be applied to any other industry vertical provided that a bilingual website exists. We also show experimental results for multilingual semantic similarity to verify the quality of the corpus and demonstrate its usage.

Hedging is a commonly used strategy in conversational management to show the speaker’s lack of commitment to what they communicate, which may signal problems between the speakers. Our project is interested in examining the presence of hedging words and phrases in identifying the tension between an interviewer and interviewee during a survivor interview. While there have been studies on hedging detection in the natural language processing literature, all existing work has focused on structured texts and formal communications. Our project thus investigated a corpus of eight unstructured conversational interviews about the Rwanda Genocide and identified hedging patterns in the interviewees’ responses. Our work produced three manually constructed lists of hedge words, booster words, and hedging phrases. Leveraging these lexicons, we developed a rule-based algorithm that detects sentence-level hedges in informal conversations such as survivor interviews. Our work also produced a dataset of 3000 sentences having the categories Hedge and Non-hedge annotated by three researchers. With experiments on this annotated dataset, we verify the efficacy of our proposed algorithm. Our work contributes to the further development of tools that identify hedges from informal conversations and discussions.

In recent NLP research, a topic of interest is universal sentence encoding, sentence representations that can be used in any supervised task.At the word sequence level, fully attention-based models suffer from two problems: a quadratic increase in memory consumption with respect to the sentence length and an inability to capture and use syntactic information.Recursive neural nets can extract very good syntactic information by traversing a tree structure.To this end, we propose Tree Transformer, a model that captures phrase level syntax for constituency trees as well as word-level dependencies for dependency trees by doing recursive traversal only with attention.Evaluation of this model on four tasks gets noteworthy results compared to the standard transformer and LSTM-based models as well as tree-structured LSTMs.Ablation studies to find whether positional information is inherently encoded in the trees and which type of attention is suitable for doing the recursive traversal are provided.

Ontologies must provide the entities, concepts, and relations required by the domain being represented. The domain of interest in this paper is the biochemistry experimental procedure. The ontology language being used is OWL-DL. OWL-DL was adopted due to its well-balanced flexibility among expressiveness (e.g., class description, cardinality restriction, etc.), completeness, and decidability. These procedures are composed of procedure steps which can be represented as sequences. Sequences are composed of totally ordered, partially ordered, and alternative subsequences. Subsequences can be represented with two relations, directlyFollows and directlyPrecedes that are used to represent sequences. Alternative subsequences can be generated by composing a oneOf function in OWL-DL, referred to it as optionalStepOf in this work, which is a simple generalization of exclusiveOR. Alkaline Agarose Gel Electrophoresis, a biochemistry procedure, is described and examples of these subsequences are provided.

Sequence labelling is the task of assigning categorical labels to a data sequence. In Natural Language Processing, sequence labelling can be applied to various fundamental problems, such as Part of Speech (POS) tagging, Named Entity Recognition (NER), and Chunking. In this study, we propose a method to adding various linguistic features to the neural sequence framework to improve sequence labelling. Besides word level knowledge, sense embeddings are added to provide semantic information. Additionally, selective readings of character embeddings are added to capture contextual as well as morphological features for each word in a sentence. Compared to previous methods, these added linguistic features allow us to design a more concise model and perform more efficient training. Our proposed architecture achieves state of the art results on the benchmark datasets of POS, NER, and chunking. Moreover, the convergence rate of our model is significantly better than the previous state of the art models.

Word sense disambiguation (WSD) is a well researched problem in computational linguistics. Different research works have approached this problem in different ways. Some state of the art results that have been achieved for this problem are by supervised models in terms of accuracy, but they often fall behind flexible knowledge-based solutions which use engineered features as well as human annotators to disambiguate every target word. This work focuses on bridging this gap using neural sequence models incorporating the well-known attention mechanism. The main gist of our work is to combine multiple attentions on different linguistic features through weights and to provide a unified framework to accomplish this. This weighted attention allows the model to easily disambiguate the sense of an ambiguous word by attending to a suitable portion of a sentence. Our extensive experiments show that weighted multiple attention enables a more versatile encoder-decoder model leading to state of the art results.

Summary sentence classification is an important step to generate document surrogates known as summary extracts. The quality of an extract depends much on the correctness of this step. We aim to classify potential summary sentences using a statistical learning method that models sentences according to a linguistic technique which examines writing styles, known as Stylometry. The sentences in documents are represented using a novel set of stylometric attributes. For learning, an innovative two-stage classification is set up that comprises two learners in subsequent steps: k-Nearest Neighbour and Naive Bayes. We train and test the learners with the newswire documents collected from two benchmark datasets, viz., the CAST and the DUC2002 datasets. Extensive experimentation strongly suggests that our method has outstanding performance for the single document summarization task. However, its performance is mixed for classifying summary sentences from multiple documents. Finally, comparisons show that our method performs significantly better than most of the popular extractive summarization methods.

Supervised machine learning methods for classifying spam emails are long-established. Most of these methods use either header-based or content-based features. Spammers, however, can bypass these methods easily-especially the ones that deal with header features. In this paper, we report a novel spam classification method that uses features based on email content-language and readability combined with the previously used content-based task features. The features are extracted from four benchmark datasets viz. CSDMC2010, Spam Assassin, Ling Spam, and Enron-Spam. We use five well-known algorithms to induce our spam classifiers: Random Forest (RF), BAGGING, ADABOOSTM1, Support Vector Machine (SVM), and Naïve Bayes (NB). We evaluate the classifier performances and find that BAGGING performs the best. Moreover, its performance surpasses that of a number of state-of-the-art methods proposed in previous studies. Although applied only to English language emails, the results indicate that our method may be an excellent means to classify spam emails in other languages, as well.

Email spam is one of the biggest threats to today's Internet. To deal with this threat, many anti-spam filters have been developed. One big challenge for these filters is to predict the labels of emails in a personalized mailbox. In this paper, we report the performance of an anti-spam filter named Sentinel. In addition to some commonplace attributes, Sentinel uses attributes related to natural language stylometry. The filter has been tested with six benchmark datasets in the Enron-Spam collection. Classifiers generated by well-known meta-learning algorithms like AdaBoostM1 and Bagging perform equally the best, while a Random Forest (RF) generated classifier performs almost as well. The performance of classifiers using Support Vector Machine (SVM) and Naive Bayes (NB) are not satisfactory. Comparisons show that the performance of Sentinel surpasses that of a number of state-of-the-art personalized filters proposed in previous studies.

Discourse models have received significant attention in the computational linguistics community with some important connections to the non-computational discourse community. More recently, the importance of discourse annotation has increased as models generated with supervised machine learning techniques are being used to annotate text automatically. A primary area for annotation is science. The theme of Teufel’s book is an important contribution in these areas: discourse models, annotation schemes, and applications.The book is a substantial work, approximately 450 pages of text and appendices. It extends Teufel’s Ph.D. thesis (Teufel 2000) with a decade of new work and updated references. The book is content-rich and meticulously written. In addition to presenting Teufel’s discourse model, it also works as a good entry point into discourse models and annotation. Because each chapter is structured with background, new material, and a summary, each chapter can be read somewhat independently. Cross-references to other parts of the book are carefully included where warranted. This structure lends itself to using the book as a reference for each of the subtopics or as an introduction to the subject area as a whole, suitable as a textbook.Chapter 1 sets the stage for the rest of the book. The author sets out her fundamental assumptions and hypotheses. The fundamental assumptions arise from three observations that she has made regarding the literature. Scientific discourse contains descriptions of positive and negative states, contains references to others’ contributions, and is the result of a rhetorical game intended to promote one’s contribution. Chapter 2, on information retrieval and citation indexes, and Chapter 3, on summarization, provide the motivation for the main theme of the book: These two information-based endeavors can be enhanced with automated tools that incorporate an understanding of the rhetorical aspects of science writing.Whereas Chapters 2 and 3 give an overview of current methodologies, Chapter 4, “New Types of Information Access,” introduces two new techniques, rhetorical extracts and citation maps, that are suggested as information navigation methods enhanced by knowledge of the discourse that contains the information being accessed. Rhetorical extracts are snippets that can be tailored to user expertise and navigation task. Citation maps are interactive citation indexes that have their citation links augmented with rhetorical or sentiment information.Chapter 5 gives a detailed description of the five scientific text corpora that are used in the research described throughout the book: computational linguistics, chemistry, genetics, cardiology, and agriculture. The chapter focuses primarily on the computational linguistics corpus, on which most of the results in the book are based. SciXML, Teufel’s markup language for science articles, is described.Chapter 6 contains an in-depth description of the Knowledge Claim Discourse Model (KCDM). Teufel gives reasons why the traditional discourse models are abandoned in favor of her new model. In addition to it being a shallow method, she points out the important aspects of KCDM (compared to Rhetorical Structure Theory): It is text-type-specific (scientific articles); no world knowledge is required; it has global (top-down) not local (bottom-up) relations; it is non-hierarchical (citation and summarization applications do not require a rich hierarchical structure).Chapter 7 presents three annotation schemes based on the KCDM: Knowledge Claim Attribution (KCA), Citation Function Classification (CFC), and Argumentative Zoning (AZ). The background and purpose of the schemes are carefully laid out. The annotation guidelines (coding manuals) are given in Appendix C.Chapter 8 reports on the reliability studies that use human annotators and gauge the quality of the annotation scheme using agreement among the annotators as a proxy for this measure. A good discussion of the measures of annotator agreement opens the chapter, followed by a detailed analysis of the four studies. Three of the four studies used three annotators, the other used 18 annotators. All studies used the computational linguistics corpus.Chapters 9 and 10 discuss the features that will be used by the machine implementations of AZ, KCA, and CFC that are described in Chapter 11. Chapter 9 provides a comprehensive discussion of the various embodiments of meta-discourse, the text that concerns itself with the dialogue between the author and the reader rather than content-bearing text. Chapter 10 discusses the computable surface features that capture the important aspects of meta-discourse that are used by the automatic annotation methods. Chapter 11 then introduces the reader to the standard supervised machine-learning methodology used to generate the statistical models that implement the automatic AZ, KCA, and CFC annotators. Chapter 12 presents gold-standard, and extrinsic and subjective evaluations of these automatic methods. The gold standard is the human-annotated computational linguistics articles and the extrinsic task is rhetorical extracts.Chapter 13 investigates the universality of the KCDM. The earlier chapters’ results were based on the computational linguistics corpus. This chapter considers the disciplines of chemistry, computer science, biology, astrophysics, and legal texts. Two issues surface: the need to modify the original KCDM slightly, and the move from an absolutely domain-knowledge-free annotation to one which includes some high-level facts about research practices in the discipline.Chapter 14 pushes the frontiers of potential uses of the KCDM methodology: support tools for scientific writing, automatic review generation, scientific summary generation that moves beyond simple sentence extraction methods and summaries of multiple scientific documents, as well as integration of automatic AZ into a large-scale digital library. Chapter 15 provides the conclusion. In the first section it recapitulates the main themes of the book. This section also nicely serves as an introduction to the book, if so desired. Section 2 lists a number of areas that could lead to an improved automatic system.The four appendices contain a list of the CmpLG-D articles, the DTD for SciXML, the annotation guidelines, and a catalog of lexical items and patterns useful in the discourse setting.The book makes an important and powerful statement in the field of discourse modeling and annotation, and provides an important body of work to which other researchers can add or compare their work. I think it is important to keep in mind the following few points while reading the book: First, Teufel comments that she is interested in a discourse model for the experimental sciences, yet her focus for much of the book is a corpus of computational linguistics papers. Also, the discourse model proposed is based on knowledge claims and rhetorical moves. This catholic view of what is science and the narrow view of structure may surprise some readers given the title of the book. Next, some of the fundamental decisions regarding the discourse model are heavily influenced by the requirements of the two motivational topics, leading one to question the full generality of the discourse model. As well, the range of rhetoric in science writing may be broader than anticipated by Teufel’s model—for example, the style found in the geology discipline is more cumulative than critical (Heather Graves, personal communication). And finally, some researchers (White 2010) argue that the domain-knowledge-free annotation dictum, although loosened slightly by Teufel, may need to be further relaxed in order to produce a more accurate gold standard, regardless of the automatic system’s access to the same domain knowledge.

Answer set programming (ASP) does not allow for incrementally constructing answer sets or locally validating constructions like proofs by only looking at a part of the given program. In this article, we elaborate upon an alternative approach to ASP that allows for incremental constructions. Our approach draws its basic intuitions from the area of default logics. We investigate the feasibility of the concept of semi-monotonicity known from default logics as a basis of incrementality. On the one hand, every logic program has at least one answer set in our alternative setting, which moreover can be constructed incrementally based on generating rules. On the other hand, the approach may produce answer sets lacking characteristic properties of standard answer sets, such as being a model of the given program. We show how integrity constraints can be used to re-establish such properties, even up to correspondence with standard answer sets. Furthermore, we develop an SLD-like proof procedure for our incremental approach to ASP, which allows for query-oriented computations. Also, we provide a characterization of our definition of answer sets via a modification of Clarks completion. Based on this notion of program completion, we present an algorithm for computing the answer sets of a logic program in our approach.

We propose and evaluate a method to identify tornadoes automatically in Doppler radar imagery by detecting hook echoes, which are important signatures of tornadoes, in Doppler radar precipitation density data. Our method uses a skeleton to represent 2D storm shapes. To characterize hook echoes, we propose four shape features of skeletons: curvature, curve orientation, thickness variation, boundary proximity, and two shape properties of tornadoes: southwest localization and the ratio of storm size to model hook echo size. To evaluate the hook echo detection algorithm, the hook echoes detected in several radar datasets by the algorithm are compared to those proposed by an expert. The effectiveness of the algorithm is quantified using a critical success index (CSI) analysis.

Amachine lifelong learningsystem based ontask rehearsalandmultiple task learning(MTL) is used to sequentially learn a series of medical diagnostic tasks. The representations of successfully learned neural network models of the tasks are stored within a domain knowledge database.Virtual examplesgenerated from these models are relearned, orrehearsed, in parallel with each new task using theηMTL neural network algorithm, a variant of MTL. TheηMTL algorithm employs a separate learning rate,ηkfor each task output,k.ηkvaries as a function of the measure of relatedness between each prior taskkand the new task being learned. Working together, the task rehearsal method andηMTL are able to develop more accurate hypotheses for a new task by selectively transferring knowledge from related tasks in domain knowledge. Coronary artery disease data sets from three real and four fictitious hospitals provide a domain of related and unrelated tasks for testing the system. The experimental results demonstrate the method's ability to sequentially retain and transfer clinical diagnostic knowledge when learning from impoverished training sets.

The availability of sequences of 3D Doppler radial velocity datasets provides sufficient information to estimate the 3D velocity of Doppler storms. We present a regularization framework for computing the 3D velocity field of storms from the underlying 3D radial velocities via an intermediate least squares computation. We obtain very realistic Doppler velocities, which can be used to estimate and predict the motion of Doppler storms. Such information is fundamental in the tracking of Doppler storms over time.

The paper presents an interactive visualization tool, Archimedean Kaleidoscope (AK), aimed at supporting a learner's exploration of polyhedra. AK uses metamorphosis as a technique to help support the learner's mental construction of relationships among different polyhedra. AK uses the symmetric nature of the platonic solids as the foundation for exploring the way in which polyhedra are related. The high level of interactivity helps support the exploration of these relationships.

A complete analysis of an English sentence includes syntactic, semantic, and pragmatic components. Presupposition belongs to the pragmatic component. How to determine the presuppositions of multiple‐clause sentences has been the focus of much work. Projection of clausal presuppositions is one method to determine the presuppositions of multiple‐clause sentences. In this paper we present a new approach to the projection problem. Drawing heavily on the theoretical techniques originating with Montague semantics, our system maps sentences of a category‐based grammar into a set of expressions of intensional logic: one expression corresponding to the literal interpretation of the sentence and the remaining expressions corresponding to the presuppositions of the sentence. The new approach correctly predicts the presuppositions of a larger range of multiple‐clause sentences than previous projection approaches.

We describe an automatic storm tracking system to help with the forecasting of severe storms. The concepts fuzzy point, fuzzy vector, fuzzy length of a fuzzy vector, and the fuzzy angle between two non-zero fuzzy vectors are first examined. We use a region splitting algorithm with dynamic thresholding to determine storm masses in Doppler radar intensity images. We represent the center of an hypothesized storm using a fuzzy point. These fuzzy storm centers are tracked over time using an incremental relaxation algorithm. The algorithms are tested on actual radar images obtained from the Atmospheric Environment Service radar station at King City, Ontario, Canada. The algorithms are capable of producing storm tracks which closely match human perception.

Forward checking (FC) is one of the most popular algorithms used to solve constraint satisfaction problems. A lazy variant of FC has been proposed called minimal forward checking (MFC). Previous empirical results suggest that MFC substantially outperforms FC when the fail first (FF) heuristic is not used. These results also suggest that the laziness of MFC can have substantial negative effects when the FF heuristic is used. To overcome this problem two extensions to the MFC algorithm are proposed, a new heuristic, called extra pruning (EXP), and the addition of conflict-directed backjumping (CBJ). An empirical investigation on a large test suite of hard randomly generated problems suggests that adding both EXP and CBJ to MFC-FF (MFC-CBJ-EXP-FF) is the best forward checking algorithm. Some theoretical relationships among the various algorithms are discussed.

An automated tracking algorithm for Doppler radar storms is presented. Potential storms in Doppler radar images are hypothesized as regions of high water density (high intensity in the radar images) using a merge-and-split region growing algorithm. Potential storms are verified by a relaxation labelling scheme that attempts to finds the best tracks based on spatio-temporal storm consistency. Temporal consistency is ensured by requiring temporal coherence of storm properties, which include size, average intensity, radial velocity variance (computed from the Doppler radial velocity images), storm shape and orientation and neighbourhooding storm disparity. Spatial consistency requires neighbouring storm tracks with common storms to compete with a winner-take-all strategy. The property coherence framework is adaptive, allowing additional properties to be added or deleted as appropriate. The tracking algorithm allows storm merging and splitting via a construction called pseudo-storms. Several tracks for Doppler storm radar data supplied by AES are given as examples of the algorithms performance.>

Forward Checking (FC) is a highly regarded complete search algorithm used to solve constraint satisfaction problems. In this paper a lazy variant of FC called minimal forward checking (MFC) is introduced. MFC is a natural marriage of incremental FC and backchecking. Given a variable selection heuristic which does not depend on domain size MPC's worst case performance on any CSP instance is the number of constraint checks performed by FC. Experiments using hard random problems are presented which show that MFC outperforms FC especially for problems with large domain sizes and/or a large number of variables.>

Some time ago, McCarthy developed the domain circumscription formalism for closed‐world reasoning. Recently, attention has been directed towards other circumscriptive formalisms. The best known of these, predicate and formula circumscription, cannot be used to produce domain‐closure axioms; nor does it appear likely that the other forms can. Since these axioms are important in deductive database theory (and elsewhere), and since domain circumscription often can conjecture these axioms, there is reason to resurrect domain circumscription.

We focus on McCarthy's method of predicate circumscription in order to establish various results about its consistency, and about its ability to conjecture new information. A basic result is that predicate circumscription cannot account for the standard kinds of default reasoning. Another is that predicate circumscription yields no new information about the equality predicate. This has important consequences for the unique names and domain closure assumptions.