So far, the task of Scientific Query-Focused Summarization (Sci-QFS) has lagged in development when compared to other areas of Scientific Natural Language Processing because of the lack of data. In this work, we propose a methodology to take advantage of existing collections of academic papers to obtain large-scale datasets for this task automatically. After applying it to the papers from our reading group, we introduce a novel dataset for Sci-QFS composed of 8,695 examples, each one with a query, the sentences of the full text from a paper and the relevance labels for each. After testing several classical and state-of-the-art embedding models on this data, we found that the task of Sci-QFS is far from being solved, although it is relatively straightforward for humans. Surprisingly, we found that classical methods outperformed modern pre-trained Deep Language Models (sometimes by a large margin), showing the need for large datasets to better fine-tune the latter. We share our experiments, data and models at https://github.com/jarobyte91/rouge_sciqfs.

In this paper, we propose a novel method to extend sequence-to-sequence models to accurately process sequences much longer than the ones used during training while being sample- and resource-efficient, supported by thorough experimentation. To investigate the effectiveness of our method, we apply it to the task of correcting documents already processed with Optical Character Recognition (OCR) systems using sequence-to-sequence models based on characters. We test our method on nine languages of the ICDAR 2019 competition on post-OCR text correction and achieve a new state-of-the-art performance in five of them. The strategy with the best performance involves splitting the input document in character n-grams and combining their individual corrections into the final output using a voting scheme that is equivalent to an ensemble of a large number of sequence models. We further investigate how to weigh the contributions from each one of the members of this ensemble. Our code for post-OCR correction is shared at https://github.com/jarobyte91/post_ocr_correction.

In this paper we describe a framework to uncover potential causal relations between event mentions from streaming text of news media. This framework relies on a dataset of manually labeled events to train a recurrent neural network for event detection. It then creates a time series of event clusters, where clusters are based on BERT contextual word embedding representations of the identified events. Using these time series dataset, we assess four methods based on Granger causality for inferring causal relations. Granger causality is a statistical concept of causality that is based on forecasting. It states that a cause occurs before the effect, and the cause produces unique changes in the effect, so past values of the cause help predict future values of the effect. The four analyzed methods are the pairwise Granger test, VAR(1), BigVar and SiMoNe. The framework is applied to the New York Times dataset, which covers news for a period of 246 months. This preliminary analysis delivers important insights into the nature of each method, identifies differences and commonalities, and points out some of their strengths and weaknesses.

Genetic Programming techniques have demonstrated great potential in dealing with the problem of query generation. This work explores different Multi-Objective Genetic Programming strategies for evolving a collection of topic-based Boolean queries. It compares three approaches to build topical Boolean queries: using terms, incorporating Wikipedia semantics (Wikipedia concepts) and a hybrid approach, using a combination of both terms and concepts. In addition, different fitness functions are combined giving rise to seven multi-objective schemes. In particular, we investigate the use of the proposed strategies in conjunction with novel fitness functions aimed at attaining high diversity based on the information-theoretic notion of entropy and Jaccard similarity. Experiments were completed using 25 topics from a dataset consisting of approximately 350,000 webpages classified into 448 topics. The results reveal that the use of Wikipedia concepts does not result in statistically significant improvements in precision, global recall or diversity when compared to the term-based approaches. However, the use of concepts has a positive effect on query interpretability since the use of terms leads to artificial queries that are hard to interpret by humans. In the meantime, concept-based queries contain a smaller number of operands than the term-based ones, hence resulting in better execution times without a loss in retrieval performance.

Electronic concept maps, interlinked with other concept maps and multimedia resources, can provide rich knowledge models to capture and share human knowledge. This article presents and evaluates methods to support experts as they extend existing knowledge models, by suggesting new context-relevant topics mined from Web search engines. The task of generating topics to support knowledge model extension raises two research questions: first, how to extract topic descriptors and discriminators from concept maps; and second, how to use these topic descriptors and discriminators to identify candidate topics on the Web with the right balance of novelty and relevance. To address these questions, this article first develops the theoretical framework required for a “topic suggester” to aid information search in the context of a knowledge model under construction. It then presents and evaluates algorithms based on this framework and applied in E xtender , an implemented tool for topic suggestion. E xtender has been developed and tested within CmapTools, a widely used system for supporting knowledge modeling using concept maps. However, the generality of the algorithms makes them applicable to a broad class of knowledge modeling systems, and to Web search in general.

Topic ontologies or web directories consist of large collections of links to websites, arranged by topic in different categories. The structure of these ontologies is typically not flat because there are hierarchical and nonhierarchical relationships among topics. As a consequence, websites classified under a certain topic may be relevant to other topics. Although some of these relevance relations are explicit, most of them must be discovered by an analysis of the structure of the ontologies. This article proposes a family of models of relevance propagation in topic ontologies. An efficient computational framework is described and used to compute nine different models for a portion of the Open Directory Project graph consisting of more than half a million nodes and approximately 1.5 million edges of different types. After performing a quantitative analysis, a user study was carried out to compare the most promising models. It was found that some general difficulties rule out the possibility of defining flawless models of relevance propagation that only take into account structural aspects of an ontology. However, there is a clear indication that including transitive relations induced by the nonhierarchical components of the ontology results in relevance propagation models that are superior to more basic approaches.

The significance of recommender systems has steadily grown in recent years as they help users to access relevant items from the vast universe of possibilities available these days. However, most of the research in recommenders is based purely on quantitative aspects, i.e., measures of similarity between items or users. In this paper we introduce a novel hybrid approach to refine recommendations achieved by quantitative methods with a qualitative approach based on argumentation, where suggestions are given after considering several arguments in favor or against the recommendations. In order to accomplish this, we use Defeasible Logic Programming (DeLP) as the underlying formalism for obtaining recommendations. This approach has a number of advantages over other existing recommendation techniques. In particular, recommendations can be refined at any time by adding new polished rules, and explanations may be provided supporting each recommendation in a way that can be easily understood by the user, by means of the computed arguments.

Abstract Formulating high‐quality queries is a key aspect of context‐based search. However, determining the effectiveness of a query is challenging because multiple objectives, such as high precision and high recall, are usually involved. In this work, we study techniques that can be applied to evolve contextualized queries when the criteria for determining query quality are based on multiple objectives. We report on the results of three different strategies for evolving queries: (a) single‐objective, (b) multiobjective with Pareto‐based ranking, and (c) multiobjective with aggregative ranking. After a comprehensive evaluation with a large set of topics, we discuss the limitations of the single‐objective approach and observe that both the Pareto‐based and aggregative strategies are highly effective for evolving topical queries. In particular, our experiments lead us to conclude that the multiobjective techniques are superior to a baseline as well as to well‐known and ad hoc query reformulation techniques.

Evolutionary algorithms rarely deal with ontogenetic, non-inherited alteration of genetic information because they are based on a direct genotype-phenotype mapping. In contrast, several processes have been discovered in nature which alter genetic information encoded in DNA before it is translated into amino-acid chains. Ontogenetically altered genetic information is not inherited but extensively used in regulation and development of phenotypes, giving organisms the ability to, in a sense, re-program their genotypes according to environmental cues. An example of post-transcriptional alteration of gene-encoding sequences is the process of RNA Editing. Here we introduce a novel Agent-based model of genotype editing and a computational study of its evolutionary performance in static and dynamic environments. This model builds on our previous Genetic Algorithm with Editing, but presents a fundamentally novel architecture in which coding and non-coding genetic components are allowed to co-evolve. Our goals are: (1) to study the role of RNA Editing regulation in the evolutionary process, (2) to understand how genotype editing leads to a different, and novel evolutionary search algorithm, and (3) the conditions under which genotype editing improves the optimization performance of traditional evolutionary algorithms. We show that genotype editing allows evolving agents to perform better in several classes of fitness functions, both in static and dynamic environments. We also present evidence that the indirect genotype/phenotype mapping resulting from genotype editing leads to a better exploration/exploitation compromise of the search process. Therefore, we show that our biologically-inspired model of genotype editing can be used to both facilitate understanding of the evolutionary role of RNA regulation based on genotype editing in biology, and advance the current state of research in Evolutionary Computation.

The understanding of the immense and intricate topological structure of the World Wide Web (WWW) is a major scientific and technological challenge. This has been recently tackled by characterizing the properties of its representative graphs, in which vertices and directed edges are identified with Web pages and hyperlinks, respectively. Data gathered in large-scale crawls have been analyzed by several groups resulting in a general picture of the WWW that encompasses many of the complex properties typical of rapidly evolving networks. In this article, we report a detailed statistical analysis of the topological properties of four different WWW graphs obtained with different crawlers. We find that, despite the very large size of the samples, the statistical measures characterizing these graphs differ quantitatively, and in some cases qualitatively, depending on the domain analyzed and the crawl used for gathering the data. This spurs the issue of the presence of sampling biases and structural differences of Web crawls that might induce properties not representative of the actual global underlying graph. In short, the stability of the widely accepted statistical description of the Web is called into question. In order to provide a more accurate characterization of the Web graph, we study statistical measures beyond the degree distribution, such as degree-degree correlation functions or the statistics of reciprocal connections. The latter appears to enclose the relevant correlations of the WWW graph and carry most of the topological information of the Web. The analysis of this quantity is also of major interest in relation to the navigability and searchability of the Web.

The last decade bears witness to an exponential growth in the use of the World Wide Web. As a result, a huge number of documents are accessible online through search engines, whose pattern-matching capabilities have turned out to be useful for mining the Web space as a particular kind of linguistic corpus, commonly known as the Web Corpus. This article presents a novel, argumentative approach to providing proactive assistance for language usage assessment on the basis of usage indices, which are good indicators of the suitability of an expression on the basis of the Web Corpus. The user preferences consist of a number of (possibly defeasible) rules and facts that encode different aspects of adequate language usage, defining the acceptability of different terms on the basis of the computed usage indices. A defeasible argumentation system determines if a given expression is ultimately acceptable by analyzing a defeasible logic program that encodes the user's preferences. © 2006 Wiley Periodicals, Inc. Int J Int Syst 21: 1151–1180, 2006.

Much intelligent user interfaces research addresses the problem of providing information relevant to a current user topic. However, little work addresses the complementary question of helping the user identify potential topics to explore next. In knowledge acquisition, this question is crucial to deciding how to extend previously-captured knowledge. This paper examines requirements for effective topic suggestion and presents a domain-independent topic-generation algorithm designed to generate candidate topics that are novel but related to the current context. The algorithm iteratively performs a cycle of topic formation, Web search for connected material, and context-based filtering. An experimental study shows that this approach significantly outperforms a baseline at developing new topics similar to those chosen by an expert for a hand-coded knowledge model.

Automatic extraction of semantic information from text and links in Web pages is key to improving the quality of search results. However, the assessment of automatic semantic measures is limited by the coverage of user studies, which do not scale with the size, heterogeneity, and growth of the Web. Here we propose to leverage human-generated metadata --- namely topical directories --- to measure semantic relationships among massive numbers of pairs of Web pages or topics. The Open Directory Project classifies millions of URLs in a topical ontology, providing a rich source from which semantic relationships between Web pages can be derived. While semantic similarity measures based on taxonomies (trees) are well studied, the design of well-founded similarity measures for objects stored in the nodes of arbitrary ontologies (graphs) is an open problem. This paper defines an information-theoretic measure of semantic similarity that exploits both the hierarchical and non-hierarchical structure of an ontology. An experimental study shows that this measure improves significantly on the traditional taxonomy-based approach. This novel measure allows us to address the general question of how text and link analyses can be combined to derive measures of relevance that are in good agreement with semantic similarity. Surprisingly, the traditional use of text similarity turns out to be ineffective for relevance ranking.

Effective knowledge management may require going beyond initial knowledge capture, to support decisions about how to extend previously-captured knowledge. Electronic concept maps, interlinked with other concept maps and multimedia resources, can provide rich knowledge models for human knowledge capture and sharing. This paper presents research on methods for supporting experts as they extend these knowledge models, by searching the Web for new context-relevant topics as candidates for inclusion. This topic search problem presents two challenges: First, how to formulate queries to seek topics that reflect the context of the current knowledge model, and, second, how to identify candidate topics with the right balance of novelty and relevance. More generally, this problem raises the broad question of the interaction of topic information from the local analysis space (a collected set of documents) and the global search space (the Web). The paper develops a framework for understanding this interaction, and proposes and evaluates techniques for addressing the query formation and topic identification questions by dynamically extracting topic descriptors and discriminators from a knowledge model, to characterize information needs for retrieval and filtering of relevant material. Using these techniques, we have developed a support tool that starts from a knowledge model under construction and automatically produces a set of suggestions for topics to include, proactively supporting users as they extend knowledge models.

An undergraduate course in Formal Languages and Automata Theory (FLAT) involves different topics that are core to the CS curricula and whose level of abstraction makes them difficult both to teach and to learn. Such difficulty stems from the complexity of the abstract notions involved and the required mathematical background. Surveys conducted among our students showed that many of them were applying some theoretical concepts mechanically rather than developing a significant learning of them, leading to a lack of motivation and interest. To cope with this problem, we introduced a number of didactic strategies based on a constructivist approach. The main aim of the proposed strategies is to promote a more significant learning of several important FLAT topics.

Electronic concept mapping tools empower experts to play an active role in the knowledge capture process, and provide a medium for building richly connected multimedia knowledge models -- sets of linked concept maps and resources about a particular domain. Knowledge models are intended to be used as a means for sharing knowledge among humans, not as carefully-crafted knowledge bases upon which machines will be performing inference. However, users must still confront the questions of what to include in a concept map and which concept maps to include in a knowledge model. This paper describes ongoing research on methods to provide content-based support to users as they extend concept maps by adding concepts and propositions, and as they select topics for new maps. The goal is to provide scaffolding for experts as they build their own concept maps, link their maps to others', and decide how to extend their knowledge models. The paper presents three approaches which start from a concept map under construction and mine related information -- both from prior concept maps, and from the web -- to propose information to aid the user's knowledge capture and knowledge construction. The paper begins with a brief summary of the concept mapping process and the CmapTools concept mapping software. It then presents three types of implemented suggesters, to suggest concepts, propositions, concept maps, and new topics to aid experts using the CmapTools, and describes preliminary experiments to assess their performance. It closes with a discussion of next steps for testing and refining these methods.