Manuel Fink


B6, 26, Room C 1.04
manuel (at)

Research Interests

■ Schema Mapping (Data Exchange)
■ Data Transformation
■ Learning from Examples

Inspired by an industry project, I am exploring techniques to learn rules for the correct transformation of large-scale databases without user interaction. In particular, given a relational database and the result of applying certain transformations to it, I am interested in reverse-engineering this process by comparing the given example data in the two different representations. This allows both the recovery of hidden transformation logic from badly documented legacy software as well as the efficient execution of multiple transformations in one step by studying their effects when applied consecutively.

My task is understood within the scientific community as the discovery of GLAV (Global-Local-As-View) schema mappings. While approaches exist already that generate mappings between heterogenous schema, a general assumption is that data values do not change. I investigate an extension of the classical problem by introducing the possibility of data transformations on top of schema transformations. For example, a geo location may not only be stored in another relation in the target schema but also with lower precision which corresponds to a truncation of digits from a geo location of the source database to produce a valid geo location in the target database.

Overlapping research fields include schema matching and instance matching whose mutual influence is amplified in my problem by the possibility of data transformations. I am also interested in relational learning in general.


Relational Learning (Exercise):			HWS 2018
Wirtschaftsinformatik IIa - Logik (Support):	FSS 2018
Künstliche Intelligenz (Exercise):		HWS 2015, HWS2016
Programmierpraktikum I (Tutorial):		HWS 2015
Programmierpraktikum II (Tutorial):		FSS 2016, FSS 2017
Programmierkurs C  (Tutorial):			FSS 2016