# before the text

Quote: Asprino, Luigi, et al. “Frame-based ontology alignment.” Thirty-First AAAI Conference on Artificial Intelligence. 2017.

The article of AAAI Summit meeting, Umm is short in length, but the detailed content can be expected.

Abstract: The need to deal with the semantic heterogeneity of resources is a key issue in the semantic Web. The most advanced ontology matching technology is the key technology to solve this problem. However, they only partially make use of the natural language descriptions of ontology entities, and most of them are impossible to find correspondence between entities with different logical types (for example, mapping properties to classes). We introduce a new approach that aims to discover the correspondence between ontology entities based on the connotation of the model, thus abstracting them from their logical types. The open data and framework semantics of links play a crucial role in this proposal. We believe that this approach may lead to technological development in the field of ontology matching and positively impact related applications, such as q&A and knowledge reconciliation.

The body of the

Introduction

An ontology is an artifact that contains a description of the domain of interest for some purpose. They provide a shared and common understanding of the domain, communicate across people and applications, and support information exchange and discovery. Due to the openness of the Web, ontologies can be defined by different people and can change quality, expressiveness, richness and coverage, thus increasing the semantic heterogeneity of resources provided through the Web of Data. Semantic heterogeneity leads to problems of redundancy and ambiguity. These issues hinder semantic interoperability between information systems and represent barriers to the development of intelligent agents that can leverage semantic information available in multiple Webource sources of knowledge.

Among the various semantic techniques for dealing with heterogeneity, ontology matching (Shvaiko and Euzenat 2013) has proven to be an effective solution for automating the integration of distributed information sources. Ontology matching (OM) discovers correspondence between semantically related entities of an ontology. However, most current ontology matching solutions present two major limitations :(I) they only partially utilize the natural language descriptions and lexical resources of ontology entities as background knowledge (for example, some examples in this direction are provided by (Giunchiglia, Shvaiko, and Yatskevich; Gracia). And Asooja)); (ii) They are mostly unable to find correspondence between entities specified by different logical types (such as mapping attributes to classes), and therefore, they are unable to deal with interpretation mismatches (as we know (Ritze et al.; Li et al. 2009)) is just an attempt to solve the problem).

Frame semantics can be used as a cognitive model to represent the formal meaning of ontology entities, thus overcoming the current limitations of current ontology matching technologies. Frame semantics (Fillmore 1982) is a formal theory of meaning based on the idea that humans can better understand the meaning of individual words by knowing knowledge related to the word. For example, the meaning of the verb to buy can be realized by knowing the circumstances in which it is applied to a business transfer involving an individual playing a particular role. Buyers, sellers, goods, money, etc. In other words, the verb buy evokes a scene in which individuals play a particular role. Our hypothesis is that a framework triggered by a word associated with an ontology entity can be used to realize the intended meaning of that entity, thereby facilitating the ontology matching task.

In this paper, we introduce a new approach aimed at discovering the correspondence between ontology entities based on the connotative meaning of the model, thus abstracting them from their logical types. We claim that comparisons between ontological entities should first be made based on the natural language associated with them, and then can be used to check for possible inconsistencies. This strategy allows us to abstract ontological entities from their lexical types by matching them with their connotations (which we believe are caused by the natural language associated with them) rather than their axiomatics. In fact, forced axiomatization can be achieved by choosing some language to specify the ontology, by the designer’s personal modeling style, or by other requirements independent of the modeling domain, such as compatibility with existing ontologies. We believe this approach may lead to a step ahead in the field of ontology matching technology and positively impact related applications such as q&A and knowledge reconciliation, ontology population and language generation.

Proposed approach

Next (Gangemi and Presutti 2010), we design an ontology matching approach that considers frames as “units of meaning” of ontologies and uses them as means of representing the conformal meaning of entities. Our strategy consists of two steps, summarized below.

Select the frame caused by the annotation. To associate ontology entities with the framework, we analyze the textual annotations associated with them. Comments provide insight into what the designer wants an entity to mean. The main idea of this approach is that the words used in the annotations evoke a framework that represents the meaning of the entity’s connotation. Our hypothesis is that the framework arising from the words contained in these annotations provides a model of the connotative meaning of the entity.

Consider the ambiguity of words when relating entities to frameworks. For example, depending on its meaning, a verb binding can evoke a frame-enforced obligation (when it is intended as “bound by obligation”) or a frame-imposed obligation (when it is intended to “surround something so as to cover or seal off”). Given this consideration, in order to associate entities with the most appropriate framework, we (I) eliminate word meanings in text-feature entities; (ii) Then, the evoked Frame is selected by using the mapping between Word Net synonyms and Frame Net’s Frame.

Mapping frameworks and ontologies. At this point, ontological entities are associated with frameworks that are somehow related (that is, evoked) to their connotations, and valid mappings between them must now be created. The integration rules for Frame Base (Rouces, de Melo, and Hose) provide examples of mappings. However, they focus on classes that convert to frames and properties into frame elements, or properties in binary projections of frames, and classes in their combined valence. The choice of some ontology types to represent concepts depends on the requirements of the extraterritorial part being represented. Therefore, we claim that mapping ontology-framework must be done without assuming any fixed correspondence between the ontology types of the two models (for example, without assuming that object attributes always correspond to binary projections of the framework).

To identify valid mappings between the ontology and the framework, we walk through the ontology entities, and for each entity, we calculate any possible mappings between the entity and the framework selected in the previous step (that is, those caused by its annotations). In framework semantics, a framework is represented by its roles (also known as frame elements), and each element may define the semantic types of individuals who can play that role within the framework. Frames, frame elements, and semantic types have names and descriptions. For each ontology entity, we calculate its similarity to the excited framework, its elements, and its semantic type. Therefore, an ontology entity can correspond to one of these components defined in the elicitation framework. This corresponding confidence is provided through the semantic textual similarity (STS) of the description of the two elements (calculated by ADW (Pilehvar, Jurgens, and Navigli)).

Frame-based ontology Matching Once the input ontology and frame are aligned, each ontology entity is associated with a formal specification of its connotations (we call it a frame-based specification). It is important to note that framework-based specifications rely on the same “language” (that is, the elements of the specification are frameworks from unique sources). In other words, the input ontology at the pointer is semantically normalized with respect to the framework. The sole purpose of this step is to compare the specification of framework-based ontology entities.

Conclusion and Future work

In this paper, we introduce a new approach for ontology matching. This approach uses frame semantics as a cognitive model to represent the connotative meaning of ontology entities. Frame-based representation makes it possible to discover the correspondence between ontology entities abstracted from their logical types, thus leading the advancement of ontology matching technology.

After the body

Has Jane recently violated the Internet Security Law? Yeah, yeah, I think Jane needs a fix, too. It’s not as creative as it was when I first came in. The operation is not very good