arXiv subjects

Manolis Koubarakis

Publications and source records attributed to Manolis Koubarakis.

18 recordsLinked to original sources

arXiv2026

AgenticER: the next frontier in Entity Resolution

Entity Resolution (ER) is a fundamental problem in data management, playing a critical role in tasks like data cleaning and knowledge graph construction. The existing ER approaches range from traditional rule-based to deep learning techniques and LLM-based methods, but typically operate under a ``passive paradigm'', as duplicates are detected through static, one-shot similarity computations. Such approaches fail to capture the inherently uncertain and context-dependent nature of real-world ER tasks, especially in data lakes with streaming content in heterogeneous formats such as CSV files, JSON files, RDF dumps, and free text. In such settings, resolving ambiguity often requires iterative evidence gathering, reasoning across multiple sources, even selective human involvement. To cover this gap, we advocate a paradigm shift from passive to Agentic ER, which frames ER as a sequential decision-making process that is performed by autonomous agents. These agents actively plan ER strategies, acquire external evidence, decide when to query additional sources or humans, and optimize trade-offs between accuracy, cost, and latency. We formalize Agentic ER as a decision-theoretic problem, we propose a reference architecture, we identify core research challenges, and outline new evaluation dimensions tailored to agentic behavior. By introducing Agentic ER, we aim to establish a new research direction at the intersection of data management and intelligent agents.

cs.DB↗

arXiv2026

DistillER: Knowledge Distillation in Entity Resolution with Large Language Models

Recent advances in Entity Resolution (ER) have leveraged Large Language Models (LLMs), achieving strong performance but at the cost of substantial computational resources or high financial overhead. Existing LLM-based ER approaches operate either in unsupervised settings and rely on very large and costly models, or in supervised settings and require ground-truth annotations, leaving a critical gap between time efficiency and effectiveness. To make LLM-powered ER more practical, we investigate Knowledge Distillation (KD) as a means to transfer knowledge from large, effective models (Teachers) to smaller, more efficient models (Students) without requiring gold labels. We introduce DistillER, the first framework that systematically bridges this gap across three dimensions: (i) Data Selection, where we study strategies for identifying informative subsets of data; (ii) Knowledge Elicitation, where we compare single- and multi-teacher settings across LLMs and smaller language models (SLMs); and (iii) Distillation Algorithms, where we evaluate supervised fine-tuning and reinforcement learning approaches. Our experiments reveal that supervised fine-tuning of Students on noisy labels generated by LLM Teachers consistently outperforms alternative KD strategies, while also enabling high-quality explanation generation. Finally, we benchmark DistillER against established supervised and unsupervised ER methods based on LLMs and SLMs, demonstrating significant improvements in both effectiveness and efficiency.

cs.DB↗

arXiv2025

Harnessing Collective Intelligence of LLMs for Robust Biomedical QA: A Multi-Model Approach

Biomedical text mining and question-answering are essential yet highly demanding tasks, particularly in the face of the exponential growth of biomedical literature. In this work, we present our participation in the 13th edition of the BioASQ challenge, which involves biomedical semantic question-answering for Task 13b and biomedical question-answering for developing topics for the Synergy task. We deploy a selection of open-source large language models (LLMs) as retrieval-augmented generators to answer biomedical questions. Various models are used to process the questions. A majority voting system combines their output to determine the final answer for Yes/No questions, while for list and factoid type questions, the union of their answers in used. We evaluated 13 state-of-the-art open source LLMs, exploring all possible model combinations to contribute to the final answer, resulting in tailored LLM pipelines for each question type. Our findings provide valuable insight into which combinations of LLMs consistently produce superior results for specific question types. In the four rounds of the 2025 BioASQ challenge, our system achieved notable results: in the Synergy task, we secured 1st place for ideal answers and 2nd place for exact answers in round 2, as well as two shared 1st places for exact answers in round 3 and 4.

cs.CL↗

arXiv2025

TerraQ: Spatiotemporal Question-Answering on Satellite Image Archives

TerraQ is a spatiotemporal question-answering engine for satellite image archives. It is a natural language processing system that is built to process requests for satellite images satisfying certain criteria. The requests can refer to image metadata and entities from a specialized knowledge base (e.g., the Emilia-Romagna region). With it, users can make requests like "Give me a hundred images of rivers near ports in France, with less than 20% snow coverage and more than 10% cloud coverage", thus making Earth Observation data more easily accessible, in-line with the current landscape of digital assistants.

cs.CV↗

arXiv2024

Transformer-based Language Models for Reasoning in the Description Logic ALCQ

Recent advancements in transformer-based language models have sparked research into their logical reasoning capabilities. Most of the benchmarks used to evaluate these models are simple: generated from short (fragments of) first-order logic sentences with only a few logical operators and quantifiers. We construct the natural language dataset, DELTA$_D$, using the expressive description logic language $\mathcal{ALCQ}$. DELTA$_D$ comprises 384K examples and increases in two dimensions: i) reasoning depth, and ii) linguistic complexity. In this way, we systematically investigate the logical reasoning capabilities of a supervised fine-tuned DeBERTa-based model and two large language models (GPT-3.5, GPT-4) with few-shot prompting. We show that the DeBERTa-based model fine-tuned on our dataset can master the entailment checking task. Moreover, the performance of GPTs can improve significantly even when a small number of samples is provided (9 shots). We open-source our code and datasets.

cs.CL↗

arXiv2024

The Large Language Model GreekLegalRoBERTa

We develop four versions of GreekLegalRoBERTa, which are four large language models trained on Greek legal and nonlegal text. We show that our models surpass the performance of GreekLegalBERT, Greek- LegalBERT-v2, and GreekBERT in two tasks involving Greek legal documents: named entity recognition and multi-class legal topic classification. We view our work as a contribution to the study of domain-specific NLP tasks in low-resource languages, like Greek, using modern NLP techniques and methodologies.

cs.CL↗

arXiv2023

Transformers in the Service of Description Logic-based Contexts

Recent advancements in transformer-based models have initiated research interests in investigating their ability to learn to perform reasoning tasks. However, most of the contexts used for this purpose are in practice very simple: generated from short (fragments of) first-order logic sentences with only a few logical operators and quantifiers. In this work, we construct the natural language dataset, DELTA$_D$, using the description logic language $\mathcal{ALCQ}$. DELTA$_D$ contains 384K examples, and increases in two dimensions: i) reasoning depth, and ii) linguistic complexity. In this way, we systematically investigate the reasoning ability of a supervised fine-tuned DeBERTa-based model and of two large language models (GPT-3.5, GPT-4) with few-shot prompting. Our results demonstrate that the DeBERTa-based model can master the reasoning task and that the performance of GPTs can improve significantly even when a small number of samples is provided (9 shots). We open-source our code and datasets.

cs.CL↗

arXiv2023

Pre-trained Embeddings for Entity Resolution: An Experimental Analysis [Experiment, Analysis & Benchmark]

Many recent works on Entity Resolution (ER) leverage Deep Learning techniques involving language models to improve effectiveness. This is applied to both main steps of ER, i.e., blocking and matching. Several pre-trained embeddings have been tested, with the most popular ones being fastText and variants of the BERT model. However, there is no detailed analysis of their pros and cons. To cover this gap, we perform a thorough experimental analysis of 12 popular language models over 17 established benchmark datasets. First, we assess their vectorization overhead for converting all input entities into dense embeddings vectors. Second, we investigate their blocking performance, performing a detailed scalability analysis, and comparing them with the state-of-the-art deep learning-based blocking method. Third, we conclude with their relative performance for both supervised and unsupervised matching. Our experimental results provide novel insights into the strengths and weaknesses of the main language models, facilitating researchers and practitioners to select the most suitable ones in practice.

cs.DB↗

arXiv2023

A Review of the Role of Causality in Developing Trustworthy AI Systems

State-of-the-art AI models largely lack an understanding of the cause-effect relationship that governs human understanding of the real world. Consequently, these models do not generalize to unseen data, often produce unfair results, and are difficult to interpret. This has led to efforts to improve the trustworthiness aspects of AI models. Recently, causal modeling and inference methods have emerged as powerful tools. This review aims to provide the reader with an overview of causal methods that have been developed to improve the trustworthiness of AI models. We hope that our contribution will motivate future research on causality-based solutions for trustworthy AI.

cs.AI↗

arXiv2022

Three-dimensional Geospatial Interlinking with JedAI-spatial

Geospatial data constitutes a considerable part of (Semantic) Web data, but so far, its sources are inadequately interlinked in the Linked Open Data cloud. Geospatial Interlinking aims to cover this gap by associating geometries with topological relations like those of the Dimensionally Extended 9-Intersection Model. Due to its quadratic time complexity, various algorithms aim to carry out Geospatial Interlinking efficiently. We present JedAI-spatial, a novel, open-source system that organizes these algorithms according to three dimensions: (i) Space Tiling, which determines the approach that reduces the search space, (ii) Budget-awareness, which distinguishes interlinking algorithms into batch and progressive ones, and (iii) Execution mode, which discerns between serial algorithms, running on a single CPU-core, and parallel ones, running on top of Apache Spark. We analytically describe JedAI-spatial's architecture and capabilities and perform thorough experiments to provide interesting insights about the relative performance of its algorithms.

cs.DB↗

arXiv2021

Multi-granular Legal Topic Classification on Greek Legislation

In this work, we study the task of classifying legal texts written in the Greek language. We introduce and make publicly available a novel dataset based on Greek legislation, consisting of more than 47 thousand official, categorized Greek legislation resources. We experiment with this dataset and evaluate a battery of advanced methods and classifiers, ranging from traditional machine learning and RNN-based methods to state-of-the-art Transformer-based methods. We show that recurrent architectures with domain-specific word embeddings offer improved overall performance while being competitive even to transformer-based models. Finally, we show that cutting-edge multilingual and monolingual transformer-based models brawl on the top of the classifiers' ranking, making us question the necessity of training monolingual transfer learning models as a rule of thumb. To the best of our knowledge, this is the first time the task of Greek legal text classification is considered in an open research project, while also Greek is a language with very limited NLP resources in general.

cs.CL↗

arXiv2021

Template-Based Question Answering over Linked Geospatial Data

Large amounts of geospatial data have been made available recently on the linked open data cloud and the portals of many national cartographic agencies (e.g., OpenStreetMap data, administrative geographies of various countries, or land cover/land use data sets). These datasets use various geospatial vocabularies and can be queried using SPARQL or its OGC-standardized extension GeoSPARQL. In this paper, we go beyond these approaches to offer a question-answering engine for natural language questions on top of linked geospatial data sources. Our system has been implemented as re-usable components of the Frankenstein question answering architecture. We give a detailed description of the system's architecture, its underlying algorithms, and its evaluation using a set of 201 natural language questions. The set of questions is offered to the research community as a gold standard dataset for the comparative evaluation of future geospatial question answering engines.

cs.IR↗

arXiv2020

OBDA for the Web: Creating Virtual RDF Graphs On Top of Web Data Sources

Due to Variety, Web data come in many different structures and formats, with HTML tables and REST APIs (e.g., social media APIs) being among the most popular ones. A big subset of Web data is also characterised by Velocity, as data gets frequently updated so that consumers can obtain the most up-to-date version of the respective datasets. At the moment, though, these data sources are not effectively supported by Semantic Web tools. To address variety and velocity, we propose Ontop4theWeb, a system that maps Web data of various formats into virtual RDF triples, thus allowing for querying them on-the-fly without materializing them as RDF. We demonstrate how Ontop4theWeb can use SPARQL to uniformly query popular, but heterogeneous Web data sources, like HTML tables and Web APIs. We showcase our approach in a number of use cases, such as Twitter, Foursquare, Yelp and HTML tables. We carried out a thorough experimental evaluation which verifies the high efficiency of our framework, which goes beyond the current state-of-the-art in this area, in terms of both functionality and performance.

cs.DB↗

arXiv2019

Evaluating Geospatial RDF stores Using the Benchmark Geographica 2

Since 2007, geospatial extensions of SPARQL, like GeoSPARQL and stSPARQL, have been defined and corresponding geospatial RDF stores have been implemented. In addition, some work on developing benchmarks for evaluating geospatial RDF stores has been carried out. In this paper, we revisit the Geographica benchmark defined by our group in 2013 which uses both real world and synthetic data to test the performance and functionality of geospatial RDF stores. We present Geographica 2, a new version of the benchmark which extends Geographica by adding one more workload, extending our existing workloads and evaluating 5 more RDF stores. Using three different real workloads, Geographica 2 tests the efficiency of primitive spatial functions in RDF stores and the performance of the RDF stores in real use case scenarios, a more detailed evaluation is performed using a synthetic workload and the scalability of the RDF stores is stressed with the scalability workload. In total eight systems are evaluated out of which six adequately support GeoSPARQL and two offer limited spatial support.

cs.DB↗

arXiv2019

Comparative Analysis of Content-based Personalized Microblog Recommendations [Experiments and Analysis]

Microblogging platforms constitute a popular means of real-time communication and information sharing. They involve such a large volume of user-generated content that their users suffer from an information deluge. To address it, numerous recommendation methods have been proposed to organize the posts a user receives according to her interests. The content-based methods typically build a text-based model for every individual user to capture her tastes and then rank the posts in her timeline according to their similarity with that model. Even though content-based methods have attracted lots of interest in the data management community, there is no comprehensive evaluation of the main factors that affect their performance. These are: (i) the representation model that converts an unstructured text into a structured representation that elucidates its characteristics, (ii) the source of the microblog posts that compose the user models, and (iii) the type of user's posting activity. To cover this gap, we systematically examine the performance of 9 state-of-the-art representation models in combination with 13 representation sources and 3 user types over a large, real dataset from Twitter comprising 60 users. We also consider a wide range of 223 plausible configurations for the representation models in order to assess their robustness with respect to their internal parameters. To facilitate the interpretation of our experimental results, we introduce a novel taxonomy of representation models. Our analysis provides novel insights into the performance and functionality of the main factors determining the performance of content-based recommendation in microblogs.

cs.IR↗

arXiv2013

Full-text Support for Publish/Subscribe Ontology Systems

We envision a publish/subscribe ontology system that is able to index millions of user subscriptions and filter them against ontology data that arrive in a streaming fashion. In this work, we propose a SPARQL extension appropriate for a publish/subscribe setting; our extension builds on the natural semantic graph matching of the language and supports the creation of full-text subscriptions. Subsequently, we propose a main-memory subscription indexing algorithm which performs both semantic and full-text matching at low complexity and minimal filtering time. Thus, when ontology data are published matching subscriptions are identified and notifications are forwarded to users.

cs.IR↗

arXiv2013

Geographica: A Benchmark for Geospatial RDF Stores

Geospatial extensions of SPARQL like GeoSPARQL and stSPARQL have recently been defined and corresponding geospatial RDF stores have been implemented. However, there is no widely used benchmark for evaluating geospatial RDF stores which takes into account recent advances to the state of the art in this area. In this paper, we develop a benchmark, called Geographica, which uses both real-world and synthetic data to test the offered functionality and the performance of some prominent geospatial RDF stores.

cs.DB↗

arXiv2012

Incomplete Information in RDF

We extend RDF with the ability to represent property values that exist, but are unknown or partially known, using constraints. Following ideas from the incomplete information literature, we develop a semantics for this extension of RDF, called RDFi, and study SPARQL query evaluation in this framework.

cs.DB↗