The paradigm of neuro-symbolic Artificial Intelligence is receiving an increasing attention in the last years to improve the results of intelligent systems by combining symbolic and subsymbolic methods. For example, existing Large Language Models (LLMs) could be enriched by taking into account background knowledge encoded using semantic technologies, such as Linguistic Linked Data (LLD). In this paper, we claim that LLD can aid Large Language Models by providing the necessary information to compute the number of poetic syllables, which would help LLMs to correctly generate poems with a valid metric. To do so, we propose an encoding for syllabic structure based on an extension of RDF vocabularies widely used in the field: POSTDATA and OntoLex-Lemon.

In this paper we introduce MUSCLE, a dataset for MUltilingual lexico-Semantic Classification of Links between Entities. The MUSCLE dataset was designed to train and evaluate Lexical Relation Classification (LRC) systems with 27K pairs of universal concepts selected from Wikidata, a large and highly multilingual factual Knowledge Graph (KG). Each pair of concepts includes its lexical forms in 25 languages and is labeled with up to five possible lexico-semantic relations between the concepts: hypernymy, hyponymy, meronymy, holonymy, and antonymy. Inspired by Semantic Map theory, the dataset bridges lexical and conceptual semantics, is more challenging and robust than previous datasets for LRC, avoids lexical memorization, is domain-balanced across entities, and enables enrichment and hierarchical information retrieval.

Extracting interesting patterns from data is the main objective of Data Mining. In this context, Frequent Itemset Mining has shown its usefulness in providing insights from transactional databases, which, in turn, can be used to gain insights about the structure of Knowledge Graphs. While there have been a lot of advances in the field, due to the NP-hard nature of the problem, the main approaches still struggle when they are faced with large databases with large and sparse vocabularies, such as the ones obtained from graph propositionalizations. There have been efforts to propose parallel algorithms, but, so far, the goal has not been to tackle this source of complexity (i.e., vocabulary size), thus, in this paper, we propose to parallelize frequent itemset mining algorithms by partitioning the database horizontally (i.e., transaction-wise) while not neglecting all the possible vertical information (i.e., item-wise). Instead of relying on pure item co-appearance metrics, we advocate for the adoption of a different approach: modeling databases as documents, where each transaction is a sentence, and each item a word. In this way, we can apply recent language modeling techniques (i.e., word embeddings) to obtain a continuous representation of the database, clusterize it in different partitions, and apply any mining algorithm to them. We show how our proposal leads to informed partitions with a reduced vocabulary size and a reduced entropy (i.e., disorder). This enhances the scalability, allowing us to speed up mining even in very large databases with sparse vocabularies. We have carried out a thorough experimental evaluation over both synthetic and real datasets showing the benefits of our proposal.

Lucia Pitarch, Jordi Bernad, Lacramioara Dranca, Carlos Bobed Lisbona, Jorge Gracia. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Public and private organizations produce and store huge amounts of documents which contain information about their domains in non-structured formats. Although from the final user’s point of view we can rely on different retrieval tools to access such data, the progressive structuring of such documents has important benefits for daily operations. While there exist many approaches to extract information in open domains, we lack tools flexible enough to adapt themselves to the particularities of different domains.

Serializable and incremental semantic reasoners make it easier to reason on a mobile device with limited resources, as they allow the reuse of previous inferences computed by another device without starting from scratch. This paper describes an extension of the fuzzy ontology reasoner fuzzyDL to make it the first serializable and incremental semantic reasoner. We empirically show that the size of the serialized files is smaller than in another serializable semantic reasoner (JFact), and that there is a significant decrease in the reasoning time.

Probabilistic Range Queries (PRQ) retrieve objects which, according to imprecise object properties, are (with a given probability) inside a precise range. When the query range is based on some imprecise object properties, which makes the query range imprecise as well, then Uncertain Probabilistic Range Queries (UPRQ) arise. Unfortunately, in the literature UPRQs ranges are constrained to be balls, i.e., the range is defined by providing a certain radius around an imprecise object property. Moreover, another important issue is the efficiency of answering UPRQs due to the necessary numerical operations to calculate probabilities. In this work we give a novel definition for UPRQs with query ranges of any shape; in addition we prove that any UPRQ can be reduced to a PRQ. Concerning the efficiency of UPRQs, we adopt and improve the usual way to address this family of queries (i.e., constructing indexes to prune/validate which objects belong to the answer, avoiding unnecessary numerical calculations) presenting: (1) a method to improve the filtering capabilities of the indexes when dealing with uniform distributions over rectangles or balls; and (2) a new index (eUD-Index), which enhances the state of the art, for any type of probability distribution. Our experiments show the feasibility of the proposals.

In this article, the authors introduce the notion of ABox intensity in the context of predicting reasoner performance to improve the representativeness of ontology metrics, and they develop new metrics that focus on ABox features of OWL 2 EL ontologies. Their experiments show that taking into account the intensity through the proposed metrics contributes to overall prediction accuracy for ABox intensive ontologies.

Many real-world applications require ontology alignment to integrate semantic information from different sources. However, most of the work in the field is restricted to finding synonymy relationships, and hyponymy relationships have not received similar attention. In this paper, we discuss some extensions of our previous work in the discovery of subsumption relationships with fuzzy clustering, aggregation operators, Formal Concept Analysis, and synonymy relationships.

This short paper motivates the need for incremental and serializable semantic reasoners. Two possible scenarios where semantic reasoners with these feature could be interesting are outlined, namely reasoning on mobile devices and managing dynamic knowledge.

We face the problem of estimating the real coverage area of a certain (mobile or static) object in a wireless scenario by considering as input data just the location of the devices that can communicate with that object. We propose and evaluate different algorithms to show that estimating the real coverage area of an object without assuming a disk of fixed radius, is not only possible but also simple and efficient.

We present a visual tool to test different algorithms that estimate the real coverage area of a certain (mobile or static) object against previously generated scenarios. Such algorithms consider as input data just the location of the devices that can communicate with that object.

In the legal field, it is a fact that a large number of documents are processed every day by management companies with the purpose of extracting data that they consider most relevant in order to be stored in their own databases. Despite technological advances, in many organizations, the task of examining these usually-extensive documents for extracting just a few essential data is still performed manually by people, which is expensive, time-consuming, and subject to human errors. Moreover, legal documents usually follow several conventions in both structure and use of language, which, while not completely formal, can be exploited to boost information extraction. In this work, we present an approach to obtain relevant information out from these legal documents based on the use of ontologies to capture and take advantage of such structure and language conventions. We have implemented our approach in a framework that allows to address different types of documents with minimal effort. Within this framework, we have also regarded one frequent problem that is found in this kind of documentation: the presence of overlapping elements, such as stamps or signatures, which greatly hinders the extraction work over scanned documents. Experimental results show promising results, showing the feasibility of our approach.

During the last years, mobile computing has been the focus of many research efforts, due mainly to the ever-growing use of mobile devices. In this context, there is a need to manage dynamic data, such as location data or other data provided by sensors. As an example, the continuous processing of location-dependent queries has been the subject of thorough research. However, there is still a need of highly expressive ways of formulating queries, augmenting in this way the systems' answer capabilities. Regarding this issue, the modeling power of Description Logics (DLs) and the inferring capabilities of their attached reasoners could fulfill this new requirement. The main problem is that DLs are inherently oriented to model static knowledge, that is, to capture the nature of the modeled objects, but not to handle changes in the property values (which requires a full ontology reclassification), as it is common in mobile computing environments (e.g., the location is expected to vary continually). In this paper, the authors present a novel approach to process continuous queries that combines 1) the DL reasoning capabilities to deal with static knowledge, with 2) the efficient data access provided by a relational database to deal with volatile knowledge. By marking at modeling time the properties that are expected to change during the lifetime of the queries, the authors'system is able to exploit both the results of the classification process provided by a DL reasoner, and the low computational costs of a database when accessing changing data (mobile environments, semantic sensors, etc.), following a two-step continuous query processing that enables us to handle continuous DL queries efficiently. Experimental results show the feasibility of the authors' approach.

Location-based services have become an increasingly interesting research area in the last two decades. However, in many scenarios, dealing with the most precise location coordinates is not the best solution since people structure the world in geographic areas instead of coordinates. Since humans work with abstractions, and names are the way we refer to those abstractions, introducing semantics in geographic definitions becomes natural. For example, users can be interested in states with vacation resorts and may want to retrieve the state names, instead of the exact geographic limits of such states. Moreover, semantics introduces new challenges, such as how to exploit the location semantics to infer new information from known definitions. For instance, we may want a system to automatically obtain the value added tax (VAT) that should be applied by a shop in Madrid, inferring the applicable tax by considering the economic area where Madrid is included (in this case, Spain); notice that the VAT should not be inferred from a bigger economic area, like Europe, although it also includes Madrid geographically. Thus, the expression of locations at different granularities extends the traditional location-based query processing to consider the most appropriate semantics for each user. In this article, adopting description logics (DLs) as a base formalism, we provide a formalization of the notion of semantic location granule and semantic granule map. We benefit from the underlying semantics of the different granularities to extend the expressivity of location-based queries and automatically discover and infer new knowledge. The model we propose uses a DL reasoner to infer new granules relationships. In particular, a DL reasoner can infer containment and intersection relationships between location granules (and help to obtain several more relationships), which provides the way to introduce semantics in location-based queries. This is done within the logical frame of DLs, thus ensuring that our approach can be supported by existing regular DL reasoners (such as Pellet, Racer Pro, and HermiT) without the need to extend their reasoning capabilities.

Nowadays, theWeb is experiencing a continuous change that is leading to the realization of the Semantic Web. Initiatives such as Linked Data have made a huge amount of structured information publicly available, encouraging the rest of the Internet community to tag their resources with it. Unfortunately, the amount of interlinked domains and information is so big that handling it efficiently has become really difficult for the final users. DBPedia, one of the biggest and most important Linked Data repositories, is a perfect example of this issue.

Nowadays, a great amount of the contents of the WWW are still mainly only human-oriented. To progressively move into the Semantic Web, the adoption of the use of ontologies is a milestone. However, their elaboration from scratch is quite expensive, which could prevent non-experts from using them in their applications. Due to this reason, different approaches for ontology reusing and engineering have been proposed, but they require domain experts and knowledge engineers. Besides, finding an ontology (if it exists) that fits a specific domain can be a tedious and frustrating experience.

The syntactic approach of most of Web search engines still has the drawback of not considering the semantics of the keywords entered by the user. So, users usually have to browse many hits looking for the information they want. In this paper, we present a system that, given a set of keywords with well defined semantics, automatically generates a set of formal queries, in the query language of the user's choice, which attempt to capture what the user had in mind when she or he wrote those keywords. The system uses ontologies and a description logics reasoner to perform a semantic enrichment of user keywords to improve the discovering of possible user queries and to reject semantically inconsistent queries.