In this paper, we explore the synergy between knowledge graph technologies and computer vision tools for personalisation systems. We propose two image user profiling approaches which map an image to knowledge graph entities representing the interests of a user who appreciates the image. The first one maps an image to entities which correspond to the objects appearing in the image. The second maps to entities which are depicted by visually similar images and which exist in the conceptual scope of the dataset within which further personalisation tasks are conducted. We show the superiority of our second approach against the baseline Google Cloud Vision API (label detection and web entity detection) in terms of accuracy metrics (precision, recall, MRR, nDCG). We also argue the importance of the capacity to create semantically useful profiles as the essence of many knowledge- or semantic-based personalisation systems is the semantic similarity calculation. We then apply our profiling approach in a novel personalisation use case where we seek to select the most appropriate images to display in recommendation banners. Our proposed knowledge-based approach tries to select the images which are the most in line with the semantic user profiles. We hypothesise that this image selection strategy allows to improve the user's perception of the recommended items. We conduct a two-stage user study with a real commercial travel dataset (1,357 package tours in 136 countries and regions depicted by 11,614 images). The results of 32 participants allow us to observe the promising performance of our approach in terms of persuasion, attention, efficiency and affinity.

In this paper, we explore the synergy between knowledge graph technologies and computer vision tools for image user profiling. We propose two image user profiling approaches which map an image to knowledge graph entities representing the interests of a user who appreciates the image. The first one maps an image to entities which correspond to the objects appearing in the image. The second one maps to entities which are depicted by visually similar images and which exist in the conceptual scope of the dataset within which further personalisation tasks are conducted. A demo configured with a real and recent commercial travel domain dataset is given at ESWC 2018.

Recommender systems are becoming must-have facilities on e-commerce websites to alleviate information overload and to improve user experience. One important component of such systems is the explanations of the recommendations. Existing explanation approaches have been classified by style and the classes are aligned with the ones for recommendation approaches, such as collaborative-based and content-based. Thanks to the semantically interconnected data, knowledge graphs have been boosting the development of content-based explanation approaches. However, most approaches focus on the exploitation of the structured semantic data to which recommended items are linked (e.g. actor, director, genre for movies). In this paper, we address the under-studied problem of leveraging knowledge graphs to explain the recommendations with items' unstructured textual description data. We point out 3 shortcomings of the state of the art entity-based explanation approach: absence of entity filtering, lack of intelligibility and poor user-friendliness. Accordingly, 3 novel approaches are proposed to alleviate these shortcomings. The first approach leverages a DBpedia category tree for filtering out incorrect and irrelevant entities. The second approach increases the intelligibility of entities with the classes of an integrated ontology (DBpedia, schema.org and YAGO). The third approach explains the recommendations with the best sentences from the textual descriptions selected by means of the entities. We showcase our approaches within a tourist tour recommendation explanation scenario and present a thorough face-to-face user study with a real commercial dataset containing 1310 tours in 106 countries. We showed the advantages of the proposed explanation approaches on five quality aspects: intelligibility, effectiveness, efficiency, relevance and satisfaction.

Résumé : L’affinité est un élément essentiel dans bien des systèmes d’information centrés sur l’utilisateur comme les systèmes de recommandation. Le graphe de connaissances et la folksonomie sont respectivement des jalons importants pour le Web Sémantique et le Web Social. Nonobstant leur trait collaboratif partagé (du moins quelques grands graphes de connaissances le sont), les données codées diffèrent tant par la nature (fait versus expérience) que par la structure (formelle versus lâche). Dans ce papier, nous tentons d’éclaircir leur performance comparative dans la tâche du calcul de l’affinité à travers deux expériences dans le domaine du etourisme. Nos résultats montrent que le graphe de connaissances permet de calculer l’affinité avec plus de précision alors que la folksonomie augmente la diversité et la nouveauté. Ces constatations nous ont motivés à développer le Framework d’Affinité Sémantique pour bénéficier de leurs avantages respectifs. L’original de ce papier est publié par ESWC 2017.

User-entity affinity is an essential component of many user-centric information systems such as online advertising, exploratory search, recommender system etc. The affinity is often assessed by analysing the interactions between users and entities within a data space. Among different affinity assessment techniques, content-based ones hypothesize that users have higher affinity with entities similar to the ones with which they had positive interactions in the past. Knowledge graph and folksonomy are respectively the milestones of Semantic Web and Social Web. Despite their shared crowdsourcing trait (not necessarily all knowledge graphs but some major large-scale ones), the encoded data are different in nature and structure. Knowledge graph encodes factual data with a formal ontology. Folksonomy encodes experience data with a loose structure. Many efforts have been made to make sense of folksonomy and to structure the community knowledge inside. Both data spaces allow to compute similarity between entities which can thereafter be used to calculate user-entity affinity. In this paper, we are interested in observing their comparative performance in the affinity assessment task. To this end, we carried out a first experiment within a travel destination recommendation scenario on a gold standard dataset. Our main findings are that knowledge graph helps to assess more accurately the affinity but folksonomy helps to increase the diversity and the novelty. This interesting complementarity motivated us to develop a semantic affinity framework to harvest the benefits of both data spaces. A second experiment with real users showed the utility of the proposed framework and confirmed our findings.

Selecting relevant travel attractions for a given user is a real and important problem from both a traveller’s and a travel supplier’s perspectives. Knowledge graphs have been used to conduct recommendations of music artists, movies and books. In this paper, we identify how knowledge graphs might be efficiently leveraged to recommend travel attractions. We improve two main drawbacks in existing systems where semantic information is exploited: semantic poorness and city-agnostic user profiling strategy. Accordingly, we constructed a rich world scale travel knowledge graph from existing large knowledge graphs namely Geonames, DBpedia and Wikidata. The underlying ontology contains more than 1200 classes to describe attractions. We applied a city-dependent user profiling strategy that makes use of the fine semantics encoded in the constructed graph. Our evaluation on YFCC100M dataset showed that our approach achieves a 5.3 % improvement in terms of F1-score, a 4.3 % improvement in terms of nDCG compared with the state-of-the-art approach.

Various studies have reported on inefficiencies of existing travel search engines, and user frustration generated through hours of searching and browsing, often with no satisfactory results. Not only do the users fail to find the right offer in the myriad of websites, but they end up browsing through many offers that do not correspond to their criteria. The Semantic Web framework is a reasonable candidate to improve this. In this paper, we present a semantic travel offer search system named "RE-ONE (Relevance Engine-One)". We especially highlight its ability to help users formulate better search queries. An example of a permitted query is in Croatia at the seaside where there is Vegetarian Restaurant. We conducted two experiments to evaluate the Query Auto-completion mechanism. The results showed that our system outperforms the Google Custom Search baseline. Queries freely conducted in RE-ONE are shown to be 63.4 % longer in terms of number of words and 27 % richer in terms of number of search criteria. RE-ONE supports better users' query formulation process by giving suggestions in greater accordance with users' idea flow.

The e-tourism is today an important field of the e-commerce. One specificity of this field is that consumers spend much time comparing many options on multiple websites before purchasing. It's easy for consumers to forget the viewed offers or websites. The Behavioral Retargeting (BR) is a widely used technique for online advertising. It leverages consumers' actions on advertisers' websites and displays relevant ads on publishers' websites. In this paper, we're interested in the relevance of the displayed ads in the e-tourism field. We present MERLOT 1, a Semantic-based travel destination recommender system that can be deployed to improve the relevance of BR in the e-tourism field. We conducted a preliminary experiment with the real data of a French travel agency. The results of 33 participants showed very promising results with regards to the baseline according to all used metrics. By this paper, we wish to provide a novel viewpoint to address the BR relevance problem, different from the dominating machine learning approaches.

In this paper we propose a new approach for improving the personaliza- tion of POIs recommender system. Existing context-aware POIs recommender sys- tems usually take into account only peripheral contextual variables. We present Ricochet, an ontology-based system that refines the recommendation results by im- plementing an inter-POI parameter that we call the "complementarity". We show how this new parameter can generate more effective recommendations. Our exper- iments are grounded using data from the location-based social network (LBSN) Yelp.com.

Dans un environnement dynamique, les ressources termino-ontologiques et les annotations sémantiques qu'elles permettent de construire doivent être modifiées régulièrement et en cohérence pour s'adapter à l'évolution du domaine sur lequel elles portent et des collections documentaires annotées. En support à un environnement d'annotation automatique de documents (TextViz), nous avons développé EvOnto pour faciliter l'évolution d'une ressource termino-ontologique en tenant compte des annotations sémantiques définies avec celle-ci. EvOnto permet de formuler une demande de changement, d'évaluer l'impact de ce changement sur la ressource termino-ontologique et sur les annotations sémantiques, et finalement de décider de la mise en oeuvre de ce changement. Cet article présente les principes d'EvOnto et une étude de cas qui illustre son apport à l'évolution d'ontologie.ABSTRACT.In dynamic environments, Ontological and Terminological Resources and semantic annotations built from them must be changed to adapt to domain evolutions and to new needs for annotated documents.Therefore, we developed the EvOnto tool that extends the TextViz tool for automatic document annotation with a termino-ontological resource.EvOnto is intended for the ontologist, it provides an interactive guide-line allowing him to formulate a change request, to evaluate its impact on the termino-ontological resource and on the semantic annotations, and finally to decide how the change will be implemented.Our paper presents the main principles of EvOnto and it reports a case-study that illustrates its support to ontology evolution.

This chapter attempts to explain the semantic Web (SW) project, which is first a vast program of enrichment of Web resources by a layer of semantic representation of content, even if this will not suffice to describe the magnitude and potential impacts of such a perspective. It discusses the real scope of SW and of different issues raised by such an approach. The chapter particularly queries the significance of the ontological attempts that oscillate between standardization of descriptions and multiple interpretations. It begins to place SW with respect to the current Web. The roles of ontologies in knowledge engineering are numerous. The chapter focuses on their specific role in relation to metadata associated with Web resources as proposed by SW. Controlled Vocabulary Terms meta data; semantic Web

In this paper we propose a method for suggesting potential collaborators for solving innovation challenges online, based on their competence, similarity of interests and social proximity with the user. We rely on Linked Data to derive a measure of semantic relatedness that we use to enrich both user profiles and innovation problems with additional relevant topics, thereby improving the performance of co-solver recommendation. We evaluate this approach against state of the art methods for query enrichment based on the distribution of topics in user profiles, and demonstrate its usefulness in recommending collaborators that are both complementary in competence and compatible with the user. Our experiments are grounded using data from the social networking service Twitter.com.

Concept recommendation is a widely used technique aimed to assist users to chose the right tags, improve their Web search experience and a multitude of other tasks. In finding potential problem solvers in Open Innovation (OI) scenarios, the concept recommendation is of a crucial importance as it can help to discover the right topics, directly or laterally related to an innovation problem. Such topics then could be used to identify relevant experts. We propose two Linked Data-based concept recommendation methods for topic discovery. The first one, hyProximity, exploits only the particularities of Linked Data structures, while the other one applies a well-known Information Retrieval method, Random Indexing, to the linked data. We compare the two methods against the baseline in the gold standard-based and user study-based evaluations, using the real problems and solutions from an OI company.

In dynamic contexts, ontologies and their lexical component (termino-ontologies or TOR) have to frequently adapt to domain evolutions, new uses and new user needs. Among all depending data, ontologybased semantic annotations also are regularly updated to annotate new documents or to reflect new points of view. Within the TextViz ontology-based annotation framework, we propose the EvOnto tool and method that supports a coherent joint change management of termino-ontologies and semantic annotations as well as quality criteria to evaluate automatic text annotations and to detect lacks in the ontology.

As more and more user traces become available as Linked Data Web, using those traces for expert finding becomes an interesting challenge, especially for the open innovation platforms. The existing expert search approaches are mostly limited to one corpus and one particular type of trace - sometimes even to a particular domain. We argue that different expert communities use different communication channels as their primary mean for communicating and disseminating knowledge, and thus different types of traces would be relevant for finding experts on different topics. We propose an approach for adapting the expert search process (choosing the right type of trace and the right expertise hypothesis) to the given topic of expertise, by relying on Linked Data metrics. In a gold standard-based experiment, we have shown that there is a significant positive correlation between the values of our metrics and the precision and recall of expert search. We also present hy.SemEx, a system that uses our Linked Data metrics to recommend the expert search approach to serve for finding experts in an open innovation scenario at hypios. The evaluation of the users' satisfaction with the system's recommendations is presented as well.

Defining roles of agents i.e., people, organisations, etc. is required in various Semantic Web applications, including access control, knowledge management and skill repository. So far, many theoretical discussions have taken place on the nature of roles and how to represent them. In this paper, we present how we implemented a lightweight OWL-DL ontology that allows to represent roles and their relations to agents. We especially focus on the various steps for choosing a well-founded model that is compatible with the general design principle of creating and consuming lightweight and easily re-usable ontology components for the Semantic Web. Our criteria to assess the “beauty” of an ontology component, in particular, are derived from practical requirements that are typical to Linked Data applications. Hence, our modeling proposal follows an original approach that bridges a gap between the Linked Data philosophy and more theoretical issues of ontology engineering. We also describe a use-case in which this ontology has been used, demonstrating in practice the benefits of our model for maintaining, browsing and querying Linked Data.

The use of semantic web technologies in enterprise environments has been subject to lots of research and development since the end of the 1990s, with topics ranging from semantic middleware infrastructures to ontology-based knowledge management systems. In this paper, we focus on the particular issue of user-driven generation of semantic web content in Enterprise 2.0 environments, and demonstrate how we extended several already-deployed user-driven applications (such as blogs, wikis and tagging systems) to produce and use semantic annotations in such context. Especially, we discuss the technical aspects involved in the process as well as the incentives that have been provided to involve users in producing such data. In addition, we present how users reacted and accepted the tools and what are the lessons that we learnt during that project.

In this paper we propose an alternative method for generating topic suggestions to be used in an Open Innovation based expert identification system. An important requirement within this challenge is the identification of topics lateral to a given innovation problem, and use them to broaden the broadcast of the problem without compromising on relevancy. We propose an approach based on DBPedia, using which we can recommend topics on certain proximity in the DBPedia concept graph. We show the important impact of the use of suggested lateral keywords to the raised awareness about the problem in a real problem broadcast.