Knowledge graph completion is still a challenging solution that uses techniques from distinct areas to solve many different tasks. Most recent works, which are based on embedding models, were conceived to improve an existing knowledge graph using the link prediction task. However, even considering the ability of these solutions to solve other tasks, they did not present results for data linking, for example. Furthermore, most of these works focuses only on structural information, i.e., the relations between entities. In this paper, we present an approach for data linking that enrich entity embeddings in a model with their literal information and that do not rely on external information of these entities. The key aspect of this proposal is that we use a blocking scheme to improve the effectiveness of the solution in relation to the use of literals. Thus, in addition to the literals from object elements in a triple, we use other literals from subjects and predicates. By merging entity embeddings with their literal information it is possible to extend many popular embedding models. Preliminary experiments were performed on real-world datasets and our solution showed competitive results to the performance of the task of data linking.
{"title":"Towards Exploring Literals to Enrich Data Linking in Knowledge Graphs","authors":"Gustavo de Assis Costa, J. P. D. Oliveira","doi":"10.1109/AIKE.2018.00024","DOIUrl":"https://doi.org/10.1109/AIKE.2018.00024","url":null,"abstract":"Knowledge graph completion is still a challenging solution that uses techniques from distinct areas to solve many different tasks. Most recent works, which are based on embedding models, were conceived to improve an existing knowledge graph using the link prediction task. However, even considering the ability of these solutions to solve other tasks, they did not present results for data linking, for example. Furthermore, most of these works focuses only on structural information, i.e., the relations between entities. In this paper, we present an approach for data linking that enrich entity embeddings in a model with their literal information and that do not rely on external information of these entities. The key aspect of this proposal is that we use a blocking scheme to improve the effectiveness of the solution in relation to the use of literals. Thus, in addition to the literals from object elements in a triple, we use other literals from subjects and predicates. By merging entity embeddings with their literal information it is possible to extend many popular embedding models. Preliminary experiments were performed on real-world datasets and our solution showed competitive results to the performance of the task of data linking.","PeriodicalId":275673,"journal":{"name":"International Conference on Artificial Intelligence and Knowledge Engineering","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123097486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The purpose of this study is to determine if the advantage of the deep learned features over the hand-crafted ones, that is evidenced in the state of the art, is still maintained for actions that are carried out in a similar environment, for real applications. The comparison is performed using a dataset created specifically for the study, in which the actions that are carried out are very similar and with a common and noisy environment. The study shows that for a database with a limited number of videos and common environment it is better to consider the hand-crafted features than a shallow CNN architecture as feature extractor.
{"title":"Deep Learned vs. Hand-Crafted Features for Action Classification","authors":"Pablo A. Arias, J. Sepúlveda","doi":"10.1109/AIKE.2018.00039","DOIUrl":"https://doi.org/10.1109/AIKE.2018.00039","url":null,"abstract":"The purpose of this study is to determine if the advantage of the deep learned features over the hand-crafted ones, that is evidenced in the state of the art, is still maintained for actions that are carried out in a similar environment, for real applications. The comparison is performed using a dataset created specifically for the study, in which the actions that are carried out are very similar and with a common and noisy environment. The study shows that for a database with a limited number of videos and common environment it is better to consider the hand-crafted features than a shallow CNN architecture as feature extractor.","PeriodicalId":275673,"journal":{"name":"International Conference on Artificial Intelligence and Knowledge Engineering","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124681075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：