Semantically-Enabled Optimization of Digital Marketing Campaigns

Abstract

Digital marketing is a domain where data analytics are a key factor to gaining competitive advantage and return of investment for companies running and monetizing digital marketing campaigns on, e.g., search engines and social media. In this paper, we propose an endto-end approach to enrich marketing campaigns performance data with third-party event data (e.g., weather events data) and to analyze the enriched data in order to predict the e↵ect of such events on campaigns' performance, with the final goal of enabling advanced optimization of the impact of digital marketing campaigns. The use of semantic technologies is central to the proposed approach: event data are made available in a format more amenable to enrichment and analytics, and the actual data enrichment technique is based on semantic data reconciliation. The enriched data are represented as Linked Data and managed in a NoSQL database to enable processing of large amounts of data. We report on the development of a pilot to build a weather-aware digital marketing campaign scheduler for JOT Internet Media-a world leading company in the digital marketing domain that has amassed a huge amount of data on campaigns performance over the years-which predicts the best date and region to launch a marketing campaign within a seven-day timespan. Additionally, we discuss benefits and limitations of applying semantic technologies to deliver better optimization strategies and competitive advantage.

Integroly: Automatic Knowledge Graph Population from Social Big Data in the Political Marketing Domain

“…The use of semantic technologies in the new digital marketing era is one of the approaches outstanding given their usefulness in decision making [30,31], data analysis [32][33][34], data integration [35][36][37], and knowledge extraction [38,39]. In decision making, semantic technologies provided the support for data integration with reasoning methods for efficiently querying the knowledge base, strategies to separate the relevant dimensions (technical, financial, political), and means to communicate the comprehensible decision options to stakeholders [30].…”

Tough Tables: Carefully Evaluating Entity Linking for Tabular Data

“…All the annotations are based on DBpedia 2016-10. 3 The structure of the dataset (depicted in Figure 2) allows the user to know Which aspects of the entity linking task are handled better/worse by different approaches. Indeed, the dataset comes with two main categories of tables:…”

“…Every day, both data practitioners and business people have to handle tables that have been extracted from databases of sales, pricing, and more. Using these tables to build a new knowledge graph (KG), populate an existing one [10], or enrich the data in the table with additional information available in existing KGs [3] requires the source data to be manipulated, interpreted within a graph-based schema (e.g., an ontology), transformed and linked to a reference or to existing KGs. The latter step, in particular, consists in an entity linking task, that is, in connecting cells to reference identifiers (e.g., URIs) that are used to describe, in larger KGs, the entity referred to the cell.…”

“…See Tables 53822652 0 5767892317858575530 and 12th Goya Awards#1 from Round 1 and Round 2, respectively. These errors come from the T2D and W2D datasets used in SemTab 2019 3. We checked our annotations against a private replica of the online DBpedia SPARQL endpoint in a local instance, loading the 2016-10 datasets listed at https://wiki.…”

Twitter-Demographer: A Flow-based Tool to Enrich Twitter Data

“…One of our goals is to provide and enforce the generation of reproducible data enrichment pipelines (i.e., they can be shared and produce the same results if components are kept the same). With data enrichment we mean the process of extending a dataset, e.g., adding new inferred properties, or disambiguating its content (Cutrona et al, 2019). Our flow-based infrastructure makes it easy to produce and share pipelines with other researchers to reconstruct the datasets.…”