A novel periodic learning ontology matching model based on interactive grasshopper optimization algorithm

Abstract

The ontology matching is a significant task for data integration and semantic interoperability. Although a large number of effective ontology matching methods have been proposed in a fully automated way, user involvement during the matching process is needed for real-world applications. It has been recognized as an effective method for further improving the quality of matching, especially for very precise matching cases. However, involving users during complex matching process suffers from new challenges of how to reduce the burden on users and how to increase effective interaction. In this paper, we propose a novel periodic learning ontology matching model based on interactive grasshopper optimization algorithm to address the above-mentioned issues. This new model takes into account the periodic feedback from users during the optimization process, rather than every generation, and a roulette wheel method is introduced to select the most problematic candidate mappings to present to users, not all, and to reduce the burden on users. To ensure the effectiveness of the interaction, a reward and punishment mechanism is considered for candidate mappings to propagate the feedback of user, and to guide the search direction of the algorithm. The experiments, conducted on two interactive tracks from Ontology Alignment Evaluation Initiative (OAEI), show that the proposed model significantly improve the quality of matching. Compared to other state-of-the-art matching systems, our model outperforms other methods in almost all cases with given different error rate, which makes it one of the most advanced leaders. Finally, a typical case of data integration is studied to present how the proposed approach is able to help enterprises to harmonize product catalogs.

Introduction

With the development of intelligent systems, ontology has been widely used in many domains and applications, for example, product lifecycle management [1], document spanning systems [2], cognitive and robotic systems [3], modern early warning system [4], intelligent transportation systems [5], and smart manufacturing systems [6], and so on. Therefore, a large number of ontologies with heterogeneity have been developed in the same domain. The problem of heterogeneity leads to the communication dilemma between application systems or humans. To solve this problem, ontology matching is a significant task for integrating heterogeneous ontologies. Therefore, many researchers have proposed ontology matching methods with different abilities [7], [8], [9], [10]. The purpose of these methods is to find a mapping set of entity pairs between different ontologies. However, the task of ontology matching remains a challenge to find high-quality alignment.

Despite many effective ontology matching systems have been developed in a fully automated way, such methods considered to have certain limitations in some knowledge domains. One of the effective methods of ontology matching is user involvement to the matching process [11], [12]. It is considered necessary in many real-world applications. Specifically, allowing the users to interactively contribute own knowledge to the mapping suggestions generated by the system during the ontology alignment process, and to further improve the quality of matching. The experiments in [13] show that user involvement is beneficial even when users make mistakes.

However, the challenge of interactive ontology matching is how to design an effective way to interact with users so that they can help to improve the quality of matching results. It is necessary and meaningful to design an interaction scheme that is not burdensome for users. Further, a good interaction design should be both natural and complete [14]. In addition, the visualization of user interaction is also a challenge. In terms of the way of user involvement, it can be divided into three categories: recommending an initial alignment in advance, selecting various matching components and weights, or providing feedback to the system during the automatic matching process. For the first category, when an initial alignment is provided as input, the matching system only needs to search a suboptimal alignment based on the initial alignment. Exactly, the user controls the behavior of the system. However, the construction of the initial alignment may impact the quality of the final alignment. The second way is to tune the strategy and parameters through user feedback. For example, the threshold tuned is used to find problematic mappings to query the user [15], [16]. The selection process of the threshold may be onerous for users, especially when increasing ontology sizes. Considering the precise solution that it suits the actual need and preferences for users, the third design way is suitable. It allows users to provide feedback on the intermediate correspondence during the matching process to improve the quality of matching. Technically, how an interactive matching system can effectively interact with users while minimizing the number of interactions is a challenge.

In recent years, meta-heuristic algorithms have been widely used in a variety of problems, such as image segmentation [2], feature selection [3], financial stress prediction [4], medical image fusion [17], and signal processing [18]. Metaheuristic technology is also known as a high-level heuristic technology or nature-inspired algorithm, which solves the problems that traditional optimization algorithms cannot solve. Recently, meta-heuristic algorithms for ontology matching have been widely concerned. Some researchers proposed ontology optimization models based on meta-heuristic algorithms, as it has demonstrated effectiveness of matching complex ontologies, which has the ability to enhance the ontology alignment process and improve the alignment quality. However, most optimization models solve ontology matching problem in a completely automatic manner. In order to incorporate user involvement, a recently proposed meta-heuristic algorithm is utilized to construct an interactive optimization model. The grasshopper optimization algorithm (GOA) is a recently proposed nature-inspired algorithm. This algorithm simulates the repulsion and attraction forces by transplanting the behavior of grasshoppers in nature. Further, the repulsion force drives grasshoppers to explore search space extensively by avoiding each other. Therefore, the algorithm has the advantage of high local optimal avoidance. The attraction force drives grasshopper swarms exploitation and convergence towards the best target. In particular, the algorithm balances exploration and exploitation by adapting the coefficient c of the comfort zone. These characteristics make the GOA algorithm more adaptable and search ability compared to other meta-heuristic algorithms in practical applications. Additionally, the interactive evolutionary computation is able to incorporate the user knowledge and preferences into the evaluation of the individual [19].

This motivated us to propose a simple and efficient ontology matching model based on interaction grasshopper optimization algorithm. The difference between interactive grasshopper optimization and non-interactive is that user feedback is used to guide the search direction.

However, new challenges arise in the optimization process of user intervention into ontology matching. That is, how to allow user involvement during the optimization process and reduce user fatigue. How to effectively interact with users to further improve the quality of matching. To address these issues, in this paper, we propose a novel periodic learning ontology matching model based on interactive grasshopper optimization algorithm. In order to reduce the burden on users, the users are required to give feedback every t generation instead of every generation. A roulette wheel technique is employed to select the most problematic candidate mappings to present to the user instead of all. In order to enhance the effectiveness of the interaction, a reward and punishment mechanism is considered to propagate the feedback of user, and to guide the search direction of swarms. The main contributions of this work are as follows:

$•$ A novel periodic learning optimization model based on interactive grasshopper optimization algorithm is proposed;

$•$ A novel roulette wheel approach is introduced into the model to select the most problematic mappings;

$•$ We proposed reward and punishment mechanisms to propagate user feedback to evolving population;

$•$ We study the effectiveness of the proposed model on two interactive tracks from Ontology Alignment Evaluation Initiative. Finally, a case is studied to further demonstrate the significance of the proposed method in practical application.

The remainder of this paper is summarized as follows. Section 2 introduces the related work for interactive ontology alignment methods. Section 3 introduces the knowledge for ontology matching problem and the original GOA briefly. Section 4 presents the periodic learning optimization model based on interaction grasshopper optimization algorithm. The time and space complexity for proposed algorithm is analyzed in Section 5. The experimental results of interactive and non-interactive are presented and analyzed in Section 6. Section 7 summarizes this paper and gives future work.