An Infrastructure for Probabilistic Reasoning with Web Ontologies

Tracking #: 990-2201

Authors: 
Jakob Huber
Mathias Niepert
Jan Noessner
Christian Meilicke
Heiner Stuckenschmidt

Responsible editor: 
Pascal Hitzler

Submission type: 
Tool/System Report
Abstract: 
We present an infrastructure for probabilistic reasoning with ontologies that is based on our Markov logic engine ROCKIT. Markov logic is a template language that combines first-order logic with log-linear graphical models. We show how to translate OWL-EL as well as RDF schema to Markov logic and how to use ROCKIT for applying MAP inference on the given set of formulas. The resulting system is an infrastructure for log linear logics that can be used for probabilistic reasoning with both extended OWL-EL and RDF schema. We describe our system and illustrate its benefits by presenting two application scenarios. These scenarios are ontology matching, and knowledge base verification, with a special focus on temporal reasoning. Our results indicate that our system, which is based on a well-founded probabilistic semantics, is capable of solving relevant problems as good as or better than state of the art systems that have specifically been designed for the respective problem.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Major Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Jie Tang submitted on 05/Apr/2015
Suggestion:
Minor Revision
Review Comment:

This manuscript was submitted as 'Tools and Systems Report' and should be reviewed along the following dimensions: (1) Quality, importance, and impact of the described tool or system (convincing evidence must be provided). (2) Clarity, illustration, and readability of the describing paper, which shall convey to the reader both the capabilities and the limitations of the tool.

The paper presents an infrastructure for probabilistic reasoning with first-order logic based on a Markov logic engine ROCKIT. It introduces the system and also describes the technical details on how to incorporate logics into the Markov model. Several applications of the proposed framework including ontology matching and knowledge-based verification have discussed and the results validate the effectiveness of the presented framework.

Personally, I enjoy reading this paper. It is well written and the organization is good. Though the paper has a favor of application on the top of an existing engine ROCKIT, the authors still give a clear description on how to define the formalize the problem into an optimization problem with logic-based constraints. In addition, the paper provides some evaluation for the presented framework using some applications such as ontology matching.

Some suggestions:
- First, regarding Markov logic network, the authors may want to cite Pedro Domingos’s work on markov logic network, who was awarded SIGKDD innovation award in 2014 because of his contribution on markov logic network. It is better to add more discussion on the difference or advantage of the presented markov logic network.
- Second, how about the efficiency of the proposed infrastructure? In fact, the first application coming to my mind based on the presented framework is semantic search. However, for search, efficiency is one key issue. It would be helpful to discuss more about efficiency.
- Third, still taking about the evaluation, I agreed that it is useful to use two applications to demonstrate the generality of the presented framework. However, each evaluation is a bit simple. My suggestion is to extend the evaluation of the ontology matching. It can be extended by either comparing with more methods or other tasks in OAEI 2013. This will make the evaluation more convincing.

Review #2
By Jing Zhang submitted on 21/Sep/2015
Suggestion:
Minor Revision
Review Comment:

This paper introduces a system that enables probabilistic reasoning in data sets described by Web ontologies. They propose to translate several Web ontology languages, such as OWL and RDF to Markov logics, and also show how to apply their previous proposed inference engine, ROCKIT, to conduct reasoning on the translated Markov logic network.

The biggest contribution of the work is the proposed method that translates the Web ontology languages to Markov logics. In addition, the demonstration of two applications, ontology mapping and knowledge base verification, solved by the proposed infrastructure, is also highly appreciated.

A few suggestions include:
1. It may be better for the authors to explain several terms or technologies more clearly as preliminaries, since some readers may not be very familiar with those terms, such as CPI, CPA, negated, unnegated, ground famula, and so on.
2. It seems that section 3 is totally about the previous work. It may be more clear if reducing the introduction of this part, while putting more efforts on introducing section 4. In my opinion, simply introducing the input, output and mechanism of ROCKIT engine is enough. Readers can refer to the original paper if they are interested in the details. How to translate Web ontology languages to Markov logics and how to apply ROCKIT on the translated Markov logics are the most important contribution of the paper and should be explained in more details.

Review #3
Anonymous submitted on 22/Sep/2015
Suggestion:
Minor Revision
Review Comment:

This manuscript was submitted as 'Tools and Systems Report' and should be reviewed along the following dimensions: (1) Quality, importance, and impact of the described tool or system (convincing evidence must be provided). (2) Clarity, illustration, and readability of the describing paper, which shall convey to the reader both the capabilities and the limitations of the tool.

The authors developed a tool to enhance probabilistic reasoning with ontologies with the focus to combing both semantic technology with machine learning approach. They all evaluated their tools using two use cases. It is still not clear on how the semantic technology can be efficiently integrated with machine learning approach to achieve the best benefits of both. This paper is written from the semantic web perspective and evaluate the systems from the semantic web point, but did not say much from the machine learning perspective. Also for the evaluation, it is not clear the size of the data, and how scalable the tool will be, what are the unique benefits of the tool comparing with other existing machine learning tools, or the combination of machine learning with logic reasoning, such as Lisa Gitoor's probabilistic soft logic (http://psl.umiacs.umd.edu/).


Comments