Abstract: Systems and computer-implemented methods are described disambiguating seller identities and assessing sellers based on continuous learning. For example, a system may identify and assess sellers across multiple channels such as electronic marketplaces by aggregating data at different time points from one or more sources of data that may each include respective and different seller details known about sellers to learn current information relating to sellers. Because of the different ways in which the sources of data may store and transmit their respective seller details, the system may aggregate the data from the different data sources, disambiguate sellers associated with the aggregated data, and generate a seller response message arranged based on a data structure that uses linkage identifiers that each groups aggregated data predicted to relate to a respective seller.