Ontology, and more broadly, Knowledge Graph Matching is a challenging task in which expressiveness has not been fully addressed. Despite the increasing use of embeddings and language models for this task, approaches for generating expressive correspondences still do not take full advantage of these models, in particular, large language models (LLMs). This paper proposes to integrate LLMs into an approach for generating expressive correspondences based on alignment need and ABox-based relation discovery. The generation of correspondences is performed by matching similar surroundings of instance sub-graphs. The integration of LLMs results in different architectural modifications, including label similarity, sub-graph matching, and entity matching. The performance word embeddings, sentence embeddings, and LLM-based embeddings, was compared. The results demonstrate that integrating LLMs surpasses all other models, enhancing the baseline version of the approach with a 45% increase in F-measure.

Existing metrics for evaluating complex ontology matching systems often fail to adequately capture the intricacies of (m:n) correspondences. This limitation results in partial or biased alignment quality assessments. This paper introduces a novel metric specifically tailored for complex ontology matching, extending traditional evaluation frameworks by incorporating subgraph similarity measures to ensure structural consistency with reference alignments. It utilizes a tree similarity-based approach, ensuring robustness against common issues such as order variance and detecting incorrect correspondences while adhering to key evaluation properties like completeness and correctness. Empirical experiments conducted on the OAEI complex track datasets demonstrate the superior adaptability of the metric in distinguishing correct structural correspondences compared to conventional and instance-based evaluation methods.

Large Language Models (LLMs) are increasingly being integrated into various components of Ontology Matching pipelines. This paper investigates the capability of LLMs to perform ontology matching directly on ontology modules and generate the corresponding alignments. Furthermore, it is explored how a dedicated fine-tuning strategy can enhance the model's matching performance in a zero-shot setting. The proposed method incorporates a search space reduction technique to select relevant subsets from both source and target ontologies, which are then used to automatically construct prompts. Recognizing the scarcity of reference alignments for training, a novel LLM-based approach is introduced for generating a synthetic dataset. This process creates a corpus of ontology submodule pairs and their corresponding reference alignments, specifically designed to fine-tune an LLM for the ontology matching task. The proposed approach was evaluated on the Conference, Geolink, Enslaved, Taxon, and Hydrography datasets from the OAEI complex track. The results demonstrate that the LLM fine-tuned on the synthetically generated data exhibits superior performance compared to the non-fine-tuned base model. The key contribution is a strategy that combines automatic dataset generation with fine-tuning to effectively adapt LLMs for ontology matching tasks.

Résumé Les approches d’alignement de propriétés de schémas de graphes de connaissances restent en retrait par rapport à la mise en correspondance des classes. Les propriétés im-pliquent souvent une variation plus importante dans leur dénomination (variation du verbe, mots fonctionnels, syno-nymes) que les classes. Cet article propose une approche d’alignement de propriétés qui combine les plongements et les extensions d’alignement afin d’améliorer les performances de la mise en correspondance de ce type d’entité. L’approche proposée est compétitive par rapport aux sys-tèmes d’alignement existants.

Matching properties still have less performance when compared to classes. Properties frequently involve a higher variation in naming (verb variation, functional words, common synonyms) than classes. Another challenge concerns the variation in property representations. This paper extends a lexical-based property matching approach combining the similarity from word and sentence pre-trained embeddings and alignment extension. The proposed approach performs competitively with state-of-the-art alignment systems on popular benchmarks in the field.

Representation learning has received increased attention in the last few years in several tasks, including knowledge graph completion, entity resolution, and ontology matching. This paper presents an overview of representation learning approaches applied to the ontology matching task. It proposes to classify such approaches into the following dimensions: lexical unit segmentation, training strategy, and information representation complexity. A discussion on them is presented together with their pros and cons. Perspectives for further developments are also discussed.

Relation Extraction (RE) is an important basic Natural Language Processing (NLP) task for many applications, including search engines and question-answering systems. There are many studies in this subarea of NLP that continue to be explored, such as the ones concerned by SemEval shared tasks. For many years, several RE systems based on statistical models have been proposed, as well as the frameworks to develop them. We focus on frameworks allowing to develop such RE systems using deep learning models. Such frameworks make it possible to reproduce experiments using many deep learning models and preprocessing techniques. Currently, there are very few frameworks of this type. In this paper, we propose an open and optimizable framework called DeepREF, inspired by two other existing frameworks: OpenNRE and REflex. DeepREF allows the rapid development of deep learning models for Relation Classification (RC). In addition, it enables hyperparameter optimization, and the application of many preprocessing techniques on the input textual data. DeepREF provides means to boost the process of running deep learning models for RC tasks on different datasets and models. DeepREF is evaluated on three reference corpora and has demonstrated competitive results compared to other state-of-the-art RC systems.

All over the world, people receive daily news on many subjects through web-based information sharing platforms such as social networks. However, some of such news are false (fake) with the potential to deceive them. Thus, the automatic detection of false news is a major issue and is gaining careful attention from the scientific community. In this paper, we present experimental analysis using both supervised and unsupervised learning on the Fake.Br corpus, a fake news dataset in Brazilian Portuguese. We propose a classification method for fake news detection based on distinct types of features, and deep learning supervised algorithms. Our best classification model achieved F1 scores up to 96% and was compared with other non-deep learning classifiers. Furthermore, we provide a complementary analysis of the same dataset by performing topic modeling based on both uni-grams and bi-grams.

Natural Language Processing (NLP) of textual data is usually broken down into a sequence of several subtasks, where the output of one the subtasks becomes the input to the following one, which constitutes an NLP pipeline. Many third-party NLP tools are currently available, each performing distinct NLP subtasks. However, it is difficult to integrate several NLP toolkits into a pipeline due to many problems, including different input/output representations or formats, distinct programming languages, and tokenization issues. This paper presents DeepNLPF, a framework that enables easy integration of third-party NLP tools, allowing the user to preprocess natural language texts at lexical, syntactic, and semantic levels. The proposed framework also provides an API for complete pipeline customization including the definition of input/output formats, integration plugin management, transparent multiprocessing execution strategies, corpus-level statistics, and database persistence. Furthermore, the DeepNLPF user-friendly GUI allows its use even by a non-expert NLP user. We conducted runtime performance analysis showing that DeepNLPF not only easily integrates existent NLP toolkits but also reduces significant runtime processing compared to executing the same NLP pipeline in a sequential manner.

This paper details the features and the methodology adopted in the construction of the CNN-corpus, a test corpus for single document extractive text summarization of news articles. The current version of the CNN-corpus encompasses 3,000 texts in English, and each of them has an abstractive and an extractive summary. The corpus allows quantitative and qualitative assessments of extractive summarization strategies.

Relation Extraction (RE) consists in detecting and classifying semantic relations between entities in a sentence.The vast majority of the state-of-the-art RE systems relies on morphosyntactic features and supervised machine learning algorithms.This paper tries to answer important questions concerning both the impact of semanticbased features, and the integration of external linguistic knowledge resources on RE performance.For that, a RE system based on a logical and relational learning algorithm was used and evaluated on three reference datasets from two distinct domains.The yielded results confirm that the classifiers induced using the proposed richer feature set outperformed the classifiers built with morphosyntactic features in average 4% (F1-measure).

A Geospatial Data Warehouse (GDW) is a repository of historical and geospatial data used in the decision-making process. These systems manage large volumes of data, and their dimensions are usually denormalized to increase query performance. Many studies have analyzed the impact of geospatial data redundancy on a relational GDW. However, to the best of our knowledge, no previous study performed a similar analysis considering the NoSQL scenario. In this context, to design a scalable document-oriented GDW (DGDW) with low storage cost and low query response time, it is important to identify which geospatial fields should be normalized (referenced) or denormalized (embedded), as well as how the documents should be partitioned among collections. In this study, we exhaustively evaluated 36 DGDWs in the MongoDB document-oriented database with different levels of geospatial redundancy and different approaches to partitioning documents among collections. Our experimental results indicate that both the normalization of low-selectivity geospatial fields and the partitioning of documents into homogenous collections provide better query performance and lower storage space. The performance evaluation presented in this paper provides strong evidence that can help guide the creation of a DGDW.

This paper proposes a topic model-based approach to extract students' weaknesses and strength based on Latent Dirichlet Allocation (LDA). Our approach combines textual data extracted from online discussion forums written by students with external sources like Wikipedia. The results show the effectiveness of the proposed approach to create a user profile based on the topics covered by the students in discussion forums.