Named Entity Matching in Publication Databases

How deep data becomes big data

“…It is actually not possible to go back and forth along the document queue to check if something similar is already loaded, especially given the potentially high cost of comparisons between structured data entities. Thus, data cleansing, deduplication and regularization (further called matching [12]) are done within RDBMS.…”

“…All algorithms that we examined in order to group instances into clusters corresponding to objects generally comprise of the following stages [5], [12]: 1) Heuristically pre-partition the set of instances onto buckets of cases which may potentially support the same objects, using some instance properties and attributes. (In some implementations, we let buckets slightly overlap with each other.)…”

“…We used SQL to extract data from particular matched clusters and processed them using some NLP methods [12], [14]. As it often happens for references, the extracted titles were modifications, sometimes quite extensive, of the original, full titles.…”

“…Key figures about the matching experiment conducted for the PMC data stored in SONCA (extension of[12])…”