=Paper=
{{Paper
|id=None
|storemode=property
|title=Types of Property Pairs and Alignment on Linked Datasets – A Preliminary Analysis
|pdfUrl=https://ceur-ws.org/Vol-1026/paper8.pdf
|volume=Vol-1026
|dblpUrl=https://dblp.org/rec/conf/i-semantics/GunaratnaTS13
}}
==Types of Property Pairs and Alignment on Linked Datasets – A Preliminary Analysis==
https://ceur-ws.org/Vol-1026/paper8.pdf
Types of Property Pairs and Alignment on Linked
Datasets – A Preliminary Analysis
Kalpa Gunaratna, Krishnaprasad Thirunarayan, and Amit Sheth
Kno.e.sis Center, Wright State University, Dayton OH, USA
{kalpa,tkprasad,amit}@knoesis.org
http://knoesis.org
Abstract. Dataset publication on the Web has been greatly influenced by the
Linked Open Data (LOD) project. Many interlinked datasets have become freely
available on the Web creating a structured and distributed knowledge represen-
tation. Analysis and aligning of concepts and instances in these interconnected
datasets have received a lot of attention in the recent past compared to properties.
We identify three different categories of property pairs found in the alignment
process and study their relative distribution among well known LOD datasets.
We also provide comparative analysis of state-of-the-art techniques with regard
to different categories, highlighting their capabilities. This could lead to more
realistic and useful alignment of properties in LOD and similar datasets.
Keywords: Linked Data, Property Alignment, Property Pair Analysis
1 Introduction
LOD [2] has popularized the way individual datasets can be published on the Web
by making inter-connections. This has resulted in the creation of a huge structured
knowledge graph on the Web. Since dataset publishers are autonomous and design
their datasets to meet their respective purposes for originally developing datasets, data
interoperability and data integration tasks on these datasets are challenging. Property
alignment is one such research problem where innovative solutions are required to han-
dle complex data representations in these interconnected datasets that go well beyond
simple string manipulations.
We introduced a novel way of computing property alignment (similarity) between
interconnected datasets by exploring the available links between the datasets and using
statistical measures [4]. Our solution can successfully handle complex data represen-
tations found at the property level in the matching process. We start with a breakdown
of types of property pairs found on the LOD and discuss the performance of matching
algorithms on the non-trivial task of property alignment between datasets. The analy-
sis is based on manually identified and categorized property pairs of a sample of well
known linked datasets in the LOD cloud. Moreover, the analysis presents how many of
the manually identified property pairs in each category are identified by the different
matching techniques (recall for each property type) highlighting their applicability.
35
Types of Property Pairs and Alignment on Linked Datasets
2 Analysis
We analyze different types of property pairs found along with the experiments per-
formed in [4]. Such analysis can provide a deeper understanding of types of property
pairs that exist in linked datasets and how matching of such property pairs can be im-
proved between two linked datasets using property extensions.
We can categorize the types of property pairs between linked datasets in two orthog-
onal ways: (1) on the basis of their semantics, and (2) on the basis of the techniques and
tools required to determine the inter-relationships or alignment among property pairs.
On the basis of semantics, the related property pairs can be classified as (1) equivalent
properties or (2) those possessing a property-sub property relationship. On the basis of
the techniques used to align properties, we can classify property pairs as follows:
1. Simple property pairs: These have high syntactic similarity in the property names
and may have a common prefix, common suffix, adjectives, or different ordering of
words, e.g., birthPlace vs placeOfBirth. Here the words “place” and “birth” are in
a different order for the two properties.
2. Opaque property pairs: These have the same meaning but use different words. This
can be further categorized into two parts.
(a) Synonymous property pairs: Similarity of the two properties can be decided
by analyzing the meaning of the property names and is intentional. This can
be achieved by using an external dictionary or a lexical database like Word-
Net. If property name is a word phrase, similarity can be checked by removing
common words from the property names, e.g., occupation vs profession, city
of birth vs place of birth. In the second property pair, the common suffix can
be eliminated from the comparison.
(b) Complex property pairs: Similarity cannot be determined by considering prop-
erty names alone, but requires additional information such as extension anal-
ysis, and domain and range. These are ambiguous or have multiple meanings
but have a specific meaning in a dataset, e.g., battle vs participated in conflict,
resting place vs place of burial. The two terms “conflict” and “resting place”
have multiple meanings and are used in many contexts. Hence, the similarity is
harder to identify.
In this analysis, we highlight the advantages of using property extensions compared
to string based and external dictionary based methods that focus on analyzing property
names in the matching process. We consider only object-type properties for this analysis
in DBpedia, Freebase, LinkedMDB, and DBLP datasets1 , taking 5000 instances in each
sample set [1]. We did not consider property chains or composite property alignment in
this preliminary analysis, which belong to the complex property pair type. Composite
property alignment is the process of aligning a property in one dataset with several prop-
erties (or property chains) in another dataset. There exist other efforts (within datasets),
different from ours, that analyze sets of properties in RDF [6], combination of proper-
ties and classes in LOD [3], and time dynamics of LOD [5].
1
person, film and software domains between DBpedia and Freebase, films between DBpedia
and LinkedMDB, and articles in DBLP (L3S and RKB Explorer).
36
Types of Property Pairs and Alignment on Linked Datasets
100% 100%
90% 90%
80% 80%
70% 70%
60% 60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
DB-FB DB-FB Film DB-FB DB-LMDB DBLP DB-FB DB-FB Film DB-FB DB-LMDB DBLP
Person Software Film Articles Person Software Film Articles
Sim Comp Syn Exact Non-Exact
(a) Property pair type breakdown (b) Exact matching of property pairs
Fig. 1. Property pairs breakdown. Syn for Synonymous, Comp for Complex and Sim for Simple
property pairs.
The correct matches in this analysis were manually identified and categorized by the
authors and verified by an external reviewer. Figure 1 shows the breakdown of proper-
ties into the three types that we are interested in. According to Figure 1(a), the majority
of the property pairs belong to simple property pairs followed by complex and syn-
onymous property pairs. Moreover, some property pairs can be matched using exact
property name matching as shown in Figure 1(b), but they account for less. Based on
the facts presented in Figure 1, on average, the majority of the matching property pairs
are simple, but cannot be matched using exact matching of property names.
There are different approaches for aligning property pairs between datasets includ-
ing [4], which is based on property extension matching. In the extension based ap-
proach, alignment of two properties is decided by aggregating the number of matched
subject-object pairs in the property extension over the number of co-appearances of
the property pair in two linked datasets. We utilized Entity Co-Reference (ECR) links
that exist between linked datasets in matching extensions. That is, two instances (in
the property extension) are considered the same if they are connected by an ECR link.
There can be incorrect matches for each property as extensions of properties overlap.
For example, ”birthPlace” property may match to ”deathPlace” with some overlap in
the extension, but when the whole result set is aggregated and analyzed, these coinsi-
dental matches can be eliminated. For the WordNet based approach, we calculated the
normalized WordNet similarity using eight similarity measures2 found in the literature
over terms appearing in the property names after removing stop words. For string sim-
ilarity measurements, we added stemming in the preprocessing step before computing
the similarity over property names. More details including threshold values and formu-
las used for matching are in [4][1].
Considering these matchers, Figure 2 shows the percentages of the correctly iden-
tified property pairs for the three types of property pairs. It also shows the superiority
of the extension based approach over string based and dictionary (WordNet) based ap-
proaches. It is clear from Figures 2(a), 2(b), 2(c), and 2(d) that the extension based
2
namely, LCH, RES, HSO, JCN, LESK, PATH, WUP and LIN
37
Types of Property Pairs and Alignment on Linked Datasets
100% 100%
90% 90%
80% 80%
70% 70%
60%
60%
50%
50% 40%
40% 30%
30% 20%
20% 10%
10% 0%
0% DB-FB DB-FB Film DB-FB DB-LMDB DBLP
Person Software Film Articles
DB-FB DB-FB Film DB-FB DB-LMDB DBLP
Person Software Film Articles Sim Sim & Exact Comp
Sim Comp Syn Comp & Exact Syn Syn & Exact
(a) Extension based algorithm matching (b) WordNet similarity based matching
100% 100%
90% 90%
80% 80%
70% 70%
60% 60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% No Results 0% No Results
DB-FB DB-FB Film DB-FB DB-LMDB DBLP DB-FB DB-FB Film DB-FB DB-LMDB DBLP
Person Software Film Articles Person Software Film Articles
Sim Comp Syn Sim Comp Syn
(c) String similarity based - Jaro Winkler (d) String similarity based - Dice
Fig. 2. Matching % (recall) for each type of property pair using different approaches. Syn for
Synonymous, Comp for Complex and Sim for Simple property pairs.
approach performed better and achieved the highest results in matching all three types
of property pairs. We added exact matching of property names capability to WordNet
based algorithm and improved its performance as shown in Figure 2(b). This is be-
cause some word phrases cannot be matched (searched) using WordNet but they have
the same or common word phrases in their names. It is also interesting to note that
the WordNet based approach failed to identify any of the synonymous property pairs
in most of the experiments as shown in Figure 2(b). This kind of behavior is expected
for string similarity or syntax based approaches, but not for a lexical database based
approach like WordNet, which is specialized in synonym word categorization. Figures
2(c) and 2(d) present matching performances when the similarity of property names are
considered using string matching algorithms. It is shown that string similarity based
matching missed all synonymous and complex property pairs leaving them unsuitable
for matching property pairs in general. Based on the facts (recall values) represented
in Figure 2, extension based property alignment has the capability to identify many
property pairs including complex and hidden property pairs compared to others. Fur-
thermore, Table 1 outlines both precision and recall for each matcher for all property
pair types, which also sheds lights on false positives (see [4] for more details). Note that
it is not possible to provide a precision value breakdown for each property pair type,
since we are not identifying each type in the alignment process but all.
38
Types of Property Pairs and Alignment on Linked Datasets
Measure DBpedia- DBpedia- DBpedia- DBpedia- DBLP RKB- Average
Type Freebase Freebase Freebase LinkedMDB DBLP L3S
(Person) (Film) (Software) (Film) (Article)
Extension Precision 0.8758 0.9737 0.6478 0.7560 1.0000 0.8427
Based Recall 0.8089* 0.5138 0.4339 0.8157 1.0000 0.7145
Algorithm F measure 0.8410* 0.6727 0.5197 0.7848 1.0000 0.7656
Precision 0.8064 0.9666 0.7659 1.0000 0.0000 0.7078
Dice
Recall 0.4777* 0.4027 0.3396 0.3421 0.0000 0.3124
Similarity
F measure 0.6000* 0.5686 0.4705 0.5098 0.0000 0.4298
Precision 0.6774 0.8809 0.7755 0.9411 0.0000 0.6550
Jaro
Recall 0.5350* 0.5138 0.3584 0.4210 0.0000 0.3656
Similarity
F measure 0.5978* 0.6491 0.4903 0.5818 0.0000 0.4638
Precision 0.5200 0.8620 0.7619 0.8823 1.0000 0.8052
WordNet
Recall 0.4140* 0.3472 0.3018 0.3947 0.3333 0.3582
Similarity
F measure 0.4609* 0.4950 0.4324 0.5454 0.5000 0.4867
Table 1. Alignment of object-type properties. Boldface and * mark highest and estimated values.
3 Conclusion
We provided a breakdown of types of property pairs that can be found on linked datasets
in the alignment process. Even though the majority of the property pairs are simple,
many cannot be identified using string manipulation techniques. In our sample datasets,
63%, 29%, and 8% of all property pairs are simple, complex, and synonymous, respec-
tively. We have shown that in every category, extension based property pair alignment
showed better results. For example, the extension based approach showed an average
improvement in the range of 5% - 32% compared to simple syntactic and WordNet
based approaches. Hence, we conclude that the extension (or instance) based approach
can discover many property pairs that are semantically the same, which cannot be un-
covered by purely syntactic means.
Acknowledgement - This work was supported by the NSF award 1143717 “III:
EAGER Expressive Scalable Querying over Linked Open Data”. Any opinions, find-
ings, and conclusions or recommendations expressed in this material are those of the
author(s) and do not necessarily reflect the views of the National Science Foundation.
References
1. More at, http://wiki.knoesis.org/index.php/Property_Alignment
2. Bizer, C., Heath, T., Berners-Lee, T.: Linked data-the story so far. International Journal on
Semantic Web and Information Systems (IJSWIS) 5(3), 1–22 (2009)
3. Gottron, T., Knauf, M., Scheglmann, S., Scherp, A.: A systematic investigation of explicit and
implicit schema information on the linked open data cloud. In: The Semantic Web: Semantics
and Big Data, pp. 228–242. Springer (2013)
4. Gunaratna, K., Thirunarayan, K., Jain, P., Sheth, A., Wijeratne, S.: A statistical and schema
independent approach to identify equivalent properties on linked data. In: 9th International
Conference on Semantic Systems. ACM (2013)
5. Käfer, T., Abdelrahman, A., Umbrich, J., OByrne, P., Hogan, A.: Observing linked data dy-
namics. In: The Semantic Web: Semantics and Big Data, pp. 213–227. Springer (2013)
6. Neumann, T., Moerkotte, G.: Characteristic sets: Accurate cardinality estimation for rdf
queries with multiple joins. In: Data Engineering (ICDE), 2011 IEEE 27th International Con-
ference on. pp. 984–994. IEEE (2011)
39