There is a need for modelling ontologies for a specialized, emerging domain, which is transnational, cultural and community studies. The ontology is synthesized as a semi-automatic model where automation exceeds to about 92 to 95% and human interference is required only to review and approve which is only about 5 to 7%. The model encompasses initial knowledge from the dataset subjected to structural topic modelling for topic synthesis and then depends upon Wikidata, CYC and NELL to obtain community contributed, community verified knowledge and formulizes into a semantic network by computing the nodal information measure. Subsequently, the knowledge is populated using terms, indices and glossaries from a knowledge stack of e-books. Classification of documents is achieved using a bagging classifier which in turn ensembles a SVM and random forest classifiers. Relevance computation as expressed by semantic similarity and reasoning is achieved using salient semantic analysis and Levenshtein’s distance with step-deviance and thresholds. The finally generated ontology is subjected to formulization, axiomatization, reasoning and finalization. Overall, precision and recall percentages of 94.04 and 96.81 respectively with FDR value of 0.06 have been achieved.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Ontology Synthesis for Transnational, Cultural and Community Studies using Hybrid Learning Paradigms

  • Alle Naga Rishikesh Reddy,
  • Gerard Deepak

摘要

There is a need for modelling ontologies for a specialized, emerging domain, which is transnational, cultural and community studies. The ontology is synthesized as a semi-automatic model where automation exceeds to about 92 to 95% and human interference is required only to review and approve which is only about 5 to 7%. The model encompasses initial knowledge from the dataset subjected to structural topic modelling for topic synthesis and then depends upon Wikidata, CYC and NELL to obtain community contributed, community verified knowledge and formulizes into a semantic network by computing the nodal information measure. Subsequently, the knowledge is populated using terms, indices and glossaries from a knowledge stack of e-books. Classification of documents is achieved using a bagging classifier which in turn ensembles a SVM and random forest classifiers. Relevance computation as expressed by semantic similarity and reasoning is achieved using salient semantic analysis and Levenshtein’s distance with step-deviance and thresholds. The finally generated ontology is subjected to formulization, axiomatization, reasoning and finalization. Overall, precision and recall percentages of 94.04 and 96.81 respectively with FDR value of 0.06 have been achieved.