Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ASPEN: ASP-Based System for Collective Entity Resolution

Zhiliang Xiang, Meghyn Bienvenu, Gianluca Cima, Víctor Gutiérrez-Basulto, Yazmín Ibáñez-García

TL;DR

ASPEN presents an ASP-based implementation of the Lace framework for collective entity resolution, addressing practical challenges in computing external similarity facts and enabling both maximal and approximate solution reasoning. By translating Lace specifications into ASP and leveraging solver features (including asprin and xclingo), ASPEN can compute, compare, and explain multiple solution sets, including LT/UT approximations and proof trees for merges. Experimental results across diverse datasets demonstrate strong accuracy, particularly in multi-relational ER, while revealing trade-offs in preprocessing and solving times, and the impactful role of recursion and DCs. Overall, ASPEN advances explainable, recursive ER via declarative programming, with viable performance and clear directions for scalability and feature expansion.

Abstract

In this paper, we present ASPEN, an answer set programming (ASP) implementation of a recently proposed declarative framework for collective entity resolution (ER). While an ASP encoding had been previously suggested, several practical issues had been neglected, most notably, the question of how to efficiently compute the (externally defined) similarity facts that are used in rule bodies. This leads us to propose new variants of the encodings (including Datalog approximations) and show how to employ different functionalities of ASP solvers to compute (maximal) solutions, and (approximations of) the sets of possible and certain merges. A comprehensive experimental evaluation of ASPEN on real-world datasets shows that the approach is promising, achieving high accuracy in real-life ER scenarios. Our experiments also yield useful insights into the relative merits of different types of (approximate) ER solutions, the impact of recursion, and factors influencing performance.

ASPEN: ASP-Based System for Collective Entity Resolution

TL;DR

ASPEN presents an ASP-based implementation of the Lace framework for collective entity resolution, addressing practical challenges in computing external similarity facts and enabling both maximal and approximate solution reasoning. By translating Lace specifications into ASP and leveraging solver features (including asprin and xclingo), ASPEN can compute, compare, and explain multiple solution sets, including LT/UT approximations and proof trees for merges. Experimental results across diverse datasets demonstrate strong accuracy, particularly in multi-relational ER, while revealing trade-offs in preprocessing and solving times, and the impactful role of recursion and DCs. Overall, ASPEN advances explainable, recursive ER via declarative programming, with viable performance and clear directions for scalability and feature expansion.

Abstract

In this paper, we present ASPEN, an answer set programming (ASP) implementation of a recently proposed declarative framework for collective entity resolution (ER). While an ASP encoding had been previously suggested, several practical issues had been neglected, most notably, the question of how to efficiently compute the (externally defined) similarity facts that are used in rule bodies. This leads us to propose new variants of the encodings (including Datalog approximations) and show how to employ different functionalities of ASP solvers to compute (maximal) solutions, and (approximations of) the sets of possible and certain merges. A comprehensive experimental evaluation of ASPEN on real-world datasets shows that the approach is promising, achieving high accuracy in real-life ER scenarios. Our experiments also yield useful insights into the relative merits of different types of (approximate) ER solutions, the impact of recursion, and factors influencing performance.
Paper Structure (27 sections, 4 theorems, 9 equations, 10 figures, 8 tables)

This paper contains 27 sections, 4 theorems, 9 equations, 10 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Entity Resolution Framework
  4. Lace Entity Resolution Specifications
  5. Semantics of Lace
  6. Summarizing and Explaining Solutions
  7. ASP Encoding and Algorithms
  8. ASP Encoding of Solutions
  9. ASP-based Algorithms
  10. Similarity Computation
  11. The ASPen System
  12. Experiments
  13. Experimental Setup
  14. Similarity Filtering
  15. Main Results
  16. ...and 12 more sections

Key Result

Theorem 1

lace_2022 It is $\mathsf{NP}$-complete (resp. $\Pi^p_2$-complete) in data complexity to decide if $\alpha \in \mathsf{PM}(D,\Sigma)$ (resp. $\mathsf{CM}(D,\Sigma)$).

Figures (10)

  • Figure 1: A schema $\mathcal{S}_\mathsf{ex}$, an $\mathcal{S}_\mathsf{ex}$-database $D_\mathsf{ex}$, and an ER specification $\Sigma_\mathsf{ex}=\langle \Gamma_\mathsf{ex}, \Delta_\mathsf{ex} \rangle$ for $\mathcal{S}_{\mathsf{ex}}$, where $\Gamma_\mathsf{ex}=\{\rho_\mathsf{ex},\sigma_\mathsf{ex}\}$ and $\Delta_\mathsf{ex}=\{\delta_\mathsf{ex}\}$.
  • Figure 2: A General Pipeline of ASPen .
  • Figure 3: Impact of Recursion on Accuracy
  • Figure 4: Merge Increments on Levels
  • Figure 5: Proof tree for merge $(s_1,s_2)$ in solution $E_1$. In the tree, $\textit{So}$ and $\textit{Ba}$ stand for the relations $\textit{Song}$ and $\textit{Band}$. $W$, $Y$, and $B$ stand for the constants Waters, 1965, and Barrett, respectively. $n_i$, $g_i$, and $t_i$ are the name and genre of the band $b_i$ and the title of song $s_i$, respectively.
  • ...and 5 more figures

Theorems & Definitions (7)

  • Example 1
  • Example 2
  • Theorem 1
  • Theorem 2
  • Example 3
  • Theorem 3: lace_2022
  • Theorem 4