arXiv subjects

David Fernandez-Amoros

Publications and source records attributed to David Fernandez-Amoros.

7 recordsLinked to original sources

arXiv2026

Reasoning About Variability Models Through Network Analysis

Feature models are widely used to capture the configuration space of software systems. Although automated reasoning has been studied for detecting problematic features and supporting configuration tasks, significantly less attention has been given to the systematic study of the structural properties of feature models at scale. The approach fills this gap by examining the models' structure through a network analysis perspective. We focus on three Research Questions concerning (i) the structural patterns exhibited by these graphs, (ii) the extent to which such patterns vary across domains and model sources, and (iii) the usefulness of network-based indicators for understanding, maintaining, and evolving variability models. To answer these questions, we analyze a dataset of 5,709 models from 20 repositories, spanning multiple application domains and varying sizes (ranging from 99 to 35,907 variables on their Boolean translation). To do so, graphs of transitive dependencies and conflicts between features are computed. Our results reveal consistent structural traits (e.g., the predominance of dependency relations, the presence of highly central features, or characteristic node degree distributions) as well as notable domain-specific deviations. These findings ease the identification of maintenance-relevant features, opportunities for modular decomposition, and indicators of structural fragility. This approach provides a scalable, graph-based foundation for the empirical analysis of variability models and contributes quantitative evidence to support future research on their structure and evolution.

cs.SE↗

arXiv2026

A Comparative Analysis of Backbone Algorithms for Configurable Software Systems

The backbone of a Boolean formula is the set of literals that must be true in every assignment that satisfies the formula. This concept is fundamental to key operations on variability models, including propagating user configuration decisions to identify implied feature selections, detecting dead features and dead code blocks, and preprocessing formulas to accelerate knowledge compilation into tractable representations such as binary decision diagrams. Despite its importance, previous empirical studies have evaluated backbone algorithms solely on SAT competition formulas (typically engineered to test the limits of SAT solvers), leading to inconsistent conclusions about their performance. This study provides the first comprehensive evaluation of formulas derived from real-world variability models, analyzing 21 configurations of 5 state-of-the-art algorithms on 2,371 formulas from configurable systems ranging from 100 variables and 179 clauses to 186,059 variables and 527,240 clauses. The results indicate that variability model formulas are structurally distinct, with higher clause density but greater clause simplicity. Our research provides clear algorithm selection guidelines: Algorithm 2/3 (iterative with solution filtering) is recommended for formulas with 1,000 or fewer variables, while Algorithm 5 (chunked core-based) with adaptive chunk size selection provides the best practical performance for larger formulas. Also, the results show that filtering heuristics have negligible or negative effects on performance for variability models. Finally, the study identifies a research gap: while Algorithm 5 with optimal chunk size can achieve runtime reductions exceeding 50\% compared to Algorithm 2/3 (the one that product line tools implement), the optimal chunk size varies unpredictably across formulas and cannot currently be estimated, opening directions for future research.

cs.SE↗

arXiv2009

Inferring Information from Feature Diagrams to Product Line Economic Models

Existing economic models support the estimation of the costs and benefits of developing and evolving a Software Product Line (SPL) as compared to undertaking traditional software development approaches. In addition, Feature Diagrams (FDs) are a valuable tool to scope the domain of a SPL. This paper proposes an algorithm to calculate, from a FD, the following information for economic models: the total number of products of a SPL, the SPL homogeneity and the commonality of the SPL requirements. The algorithm running time belongs to the complexity class $O(f^42^c)$. In contrast to related work, the algorithm is free of dependencies on off-the-self tools and is generally specified for an abstract FD notation, that works as a pivot language for most of the available notations for feature modeling.

cs.SE↗

arXiv2009

Word Sense Disambiguation Using English-Spanish Aligned Phrases over Comparable Corpora

In this paper we describe a WSD experiment based on bilingual English-Spanish comparable corpora in which individual noun phrases have been identified and aligned with their respective counterparts in the other language. The evaluation of the experiment has been carried out against SemCor. We show that, with the alignment algorithm employed, potential precision is high (74.3%), however the coverage of the method is low (2.7%), due to alignments being far less frequent than we expected. Contrary to our intuition, precision does not rise consistently with the number of alignments. The coverage is low due to several factors; there are important domain differences, and English and Spanish are too close languages for this approach to be able to discriminate efficiently between senses, rendering it unsuitable for WSD, although the method may prove more productive in machine translation.

cs.CL↗

arXiv2009

The Uned systems at Senseval-2

We have participated in the SENSEVAL-2 English tasks (all words and lexical sample) with an unsupervised system based on mutual information measured over a large corpus (277 million words) and some additional heuristics. A supervised extension of the system was also presented to the lexical sample task. Our system scored first among unsupervised systems in both tasks: 56.9% recall in all words, 40.2% in lexical sample. This is slightly worse than the first sense heuristic for all words and 3.6% better for the lexical sample, a strong indication that unsupervised Word Sense Disambiguation remains being a strong challenge.

cs.CL↗

arXiv2009

Word Sense Disambiguation Based on Mutual Information and Syntactic Patterns

This paper describes a hybrid system for WSD, presented to the English all-words and lexical-sample tasks, that relies on two different unsupervised approaches. The first one selects the senses according to mutual information proximity between a context word a variant of the sense. The second heuristic analyzes the examples of use in the glosses of the senses so that simple syntactic patterns are inferred. This patterns are matched against the disambiguation contexts. We show that the first heuristic obtains a precision and recall of .58 and .35 respectively in the all words task while the second obtains .80 and .25. The high precision obtained recommends deeper research of the techniques. Results for the lexical sample task are also provided.

cs.CL↗

arXiv2001

The Role of Conceptual Relations in Word Sense Disambiguation

We explore many ways of using conceptual distance measures in Word Sense Disambiguation, starting with the Agirre-Rigau conceptual density measure. We use a generalized form of this measure, introducing many (parameterized) refinements and performing an exhaustive evaluation of all meaningful combinations. We finally obtain a 42% improvement over the original algorithm, and show that measures of conceptual distance are not worse indicators for sense disambiguation than measures based on word-coocurrence (exemplified by the Lesk algorithm). Our results, however, reinforce the idea that only a combination of different sources of knowledge might eventually lead to accurate word sense disambiguation.

cs.CL↗