TY - GEN
T1 - DataXFormer
T2 - 32nd IEEE International Conference on Data Engineering, ICDE 2016
AU - Abedjan, Ziawasch
AU - Morcos, John
AU - Ilyas, Ihab F.
AU - Ouzzani, Mourad
AU - Papotti, Paolo
AU - Stonebraker, Michael
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2016/6/22
Y1 - 2016/6/22
N2 - In data integration, data curation, and other data analysis tasks, users spend a considerable amount of time converting data from one representation to another. For example US dates to European dates or airport codes to city names. In a previous vision paper, we presented the initial design of DataXFormer, a system that uses web resources to assist in transformation discovery. Specifically, DataXFormer discovers possible transformations from web tables and web forms and involves human feedback where appropriate. In this paper, we present the full fledged system along with several extensions. In particular, we present algorithms to find (i) transformations that entail multiple columns of input data, (ii) indirect transformations that are compositions of other transformations, (iii) transformations that are not functions but rather relationships, and (iv) transformations from a knowledge base of public data. We report on experiments with a collection of 120 transformation tasks, and show our enhanced system automatically covers 101 of them by using openly available resources.
AB - In data integration, data curation, and other data analysis tasks, users spend a considerable amount of time converting data from one representation to another. For example US dates to European dates or airport codes to city names. In a previous vision paper, we presented the initial design of DataXFormer, a system that uses web resources to assist in transformation discovery. Specifically, DataXFormer discovers possible transformations from web tables and web forms and involves human feedback where appropriate. In this paper, we present the full fledged system along with several extensions. In particular, we present algorithms to find (i) transformations that entail multiple columns of input data, (ii) indirect transformations that are compositions of other transformations, (iii) transformations that are not functions but rather relationships, and (iv) transformations from a knowledge base of public data. We report on experiments with a collection of 120 transformation tasks, and show our enhanced system automatically covers 101 of them by using openly available resources.
UR - http://www.scopus.com/inward/record.url?scp=84980367317&partnerID=8YFLogxK
U2 - 10.1109/ICDE.2016.7498319
DO - 10.1109/ICDE.2016.7498319
M3 - Conference contribution
AN - SCOPUS:84980367317
T3 - 2016 IEEE 32nd International Conference on Data Engineering, ICDE 2016
SP - 1134
EP - 1145
BT - 2016 IEEE 32nd International Conference on Data Engineering, ICDE 2016
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 16 May 2016 through 20 May 2016
ER -