Integrolysistema automático de población de un grafo de conocimiento a partir de datos sociales masivos en el dominio de marketing político

Abstract

Modern political campaigns are based on strategies formulated in what has become known as political marketing. Political marketing ranges from the definition of the political product, through a detailed analysis of the needs of the electorate, to the development of campaigns and the management of political communication. Market intelligence is understanding the real demands of the political market, will be the first stage on this complex process. One of the main problems of political market intelligence is the processing of the diversity of data sources such as surveys, social networking sites, Web pages, databases of political institutions, among others. Social networks sites have become one of the main conversation platforms and channels for sharing experiences and opinions. They encourage public discourse and, in particular, are increasingly used for political issues, such as citizen participation, proselytism or political debate. However, the monitoring, extraction, processing, storage and analysis of social big data in the political sphere is a complicated task due to the heterogeneity of the data source structure. Despite the challenges and complications, these tasks are very important to meet the short, medium and long-term demands of potential voters, and thereby create a strategy and decision-making linked to them. Semantic Web technologies, and in particular, ontologies, allow the generation of a homogeneous knowledge base, shared and reused, with a machine-readable format and enriched to support the discovery, integration, representation and management of knowledge. In recent years, ontologies and Knowledge Graphs (KGs) have proven to be effective in enabling a data integration solution and helping with the complexity of finding meaningful relationships within heterogeneous data. This work focuses on two main contributions, the first, to generate an ontological model based on the domain of political marketing, and the second, in a framework proposal for automatic Knowledge Graph population from social big data in the political marketing domain. Our ontological model proposal was made based on the criteria and needs of political marketing on political campaigns. The construction of the ontology was through the standard Ontology 101 methodology, where the domain and scope of the ontology were defined, the reuse of existing ontologies was considered, the important terms for the ontology were listed, up to the definition of classes, properties, relationships and base instances. The ontology was named PMont (Political Marketing Ontology) which answers key questions of political marketing. The PMont allowed the integration of the information available on the electorate and candidates through different data sources. The second contribution of this research work is an automated system based on texts in Spanish through Machine Learning (ML) and Natural Language Processing (NLP) techniques, collecting significant data from semi-structured and unstructured digital media sources, processing big data and finally the population of a Knowledge Graph previously defined by an ontological model of the political marketing domain. The automation system proposal has the following phases or components: (i) source connection and data collection, (ii) information extraction, (iii) ontology population and validation process, and finally, (iv) the Knowledge Graph. While the process is carried out automatically, the instances added to the Knowledge Graph are annotated so that human experts can supervise the results of the automated population in an optional later stage. As a result of this doctoral thesis, we have a Knowledge Graph populated with validated, precise, consistent, and reliable information. The Knowledge Graph was evaluated through 18 quality requirements on the dimensions of accessibility, contextuality, intrinsic and representativeness, of which it provides us with an optimal knowledge base for political marketing market intelligence.