Paper 128 (Research track)

A Parameterized Formal Model for Flexibly Defining Schema-level Indices for the Web of Data

Author(s): Till Blume, Matthias Schulte-Althoff, Thomas Gottron, Ansgar Scherp

Abstract: Schema-level indices address many challenges that come with the growing size of the Web of Data. We argue that there is not a one-size-fits-all index for the Web of Data. Rather, a parameterized, formal model is needed that allows to quickly design, tailor, and compare different schema-level indices. We provide the first formal model for schema-level indices called FLuID. Our model is abstracted from existing indices. In addition, our model provides novel features such as aggregation over owl:sameAs as well as RDFS inferencing. Indices defined with FLuID can be efficiently computed in O(n). We implemented the FLuID model following an existing stream-based schema computation approach for the Web of Data. We empirically show that indeed different index models are needed for different information needs, datasets, and space requirements.

Keywords: Linked data; Schema-level indices; Parameterized meta model

Share on

One thought to “Paper 128 (Research track)”

  1. I like the paper.

    Several remarks:
    Not all structural summaries use bisimulation. E.g. ABSTAT patterns do not correspond to equivalence classes on individuals, as individuals might have belong to more patterns (e.g. (E,Q,D),(C,Q,D) in Fig.1. of [1]) – if you use the ABSTAT index differently, it should be mentioned in the paper in my view. Similarly, note that in SchemEx only two of the three layers are partitions (TC, EQC). The first layer is not (an instance can belong to more clusters).

    p4 – “ABSTAT selects one type …”: ABSTAT does not seem to select a single type – instead [1] assumes multiple minimal types – line 3 of page 5 of [1].
    p4 – “Another example is …” : SchemEx uses bisimulation, while ABSTAT does not
    p7 – “The two instances [i2]_I and [i3]_I …” : should be in my view “The two instances [i3]_I and [i4]_I …”

    [1] ABSTAT: Ontology-driven Linked Data Summaries with Pattern Minimization

Leave a Reply

Your email address will not be published. Required fields are marked *