An overview about data integration in data lakes

Integrating data in data lakes is essential so we can perform more complex analyses. However, data lakes are mainly composed of raw data, from structured, semi-structured, and even unstructured data. It turns out that traditional data integration algorithms usually expect to receive structured data as input, so those different types of data jeopardize big data integration. This paper presents a systematic literature review that generates a broad landscape about data integration in data lakes. We searched for papers in eight well-known search engines, following a structured process. From the 298 papers we retrieved, we selected 22 papers that answer our research questions. We identify examples of data lake integration models, how they calculate the similarity among the datasets, how the models are evaluated, the most common type of data they integrate, and the challenges inherent to the area, which points to future research directions in data integration in data lakes.