The implementation of SchemaRank proposed in ISWC'21. If you use our software, please cite the following paper.

Hiroaki Shiokawa,  
"Fast ObjectRank for Large Knowledge Databases,"
Proceedings of the 20th International Semantic Web Conference (ISWC2021), pp.217-234, October, 2021.

@inproceedings{Shiokawa2021b,
  title   = {{Fast ObjectRank for Large Knowledge Databases}},
  author  = {Hiroaki Shiokawa},
  booktitle = {The Semantic Web -- ISWC 2021},
  pages   = {217--234},
  year    = {2021},
}

This software is released under the MIT license. Please read LICENSE for details.

SchemaRank requires the following softwares.

• gcc Version 9.2.0 (or later)

We have confirmed that our software works on the following environments.

• CentOS 8
• MacOS 11.6

1. Install g++ on your environment.
2. Build the softwares in ./generate_crs and ./.

$ cd generate_crs
$ make clean; make
$ cd ..
$ make clean; make

3. If you can find schemarank and gen_crs in ./ and ./generate_crs, respectively, the build has been successfully completed.

SchemaRank requires three input files: schema_graph.txt, data_graph.txt, and query.txt.

schema_graph.txt is a file that represents a schema graph of a given knowledge graph. The file must include a list of adjacent node-types in the schema graph with a corresponding edge-weight value like as follows:

A	Ps	0.2
Ps	A	0.2
Ps	Pd	0.7
Ps	Y	0.1
Y	Ps	0.3
Y	C	0.3
C	Y	0.3

For example, if node-types A and Ps are connected with a weighted edge with 0.2, schema_graph.txt should have a line of A\tPs\t0.2, where \t is TAB.

data_graph is a file that represents a data graph of a given knowledge graph. The file must include a list of adjacent nodes in the data graph along with a corresponding edge weitht value and node-types.

0	1	0.15	CY
0	2	0.15	CY
1	0	0.3	YC
1	3	0.3	YPs
2	0	0.3	YC
2	4	0.3	YPs
3	1	0.1	PsY
3	4	0.7	PsPd
3	5	0.2	PsA
4	2	0.1	PsY
4	5	0.1	PsA
4	6	0.1	PsA
5	3	0.1	APs
5	4	0.1	APs
6	4	0.2	APs

For example, if a data graph has an edge between node-id 0 and 1 with a weight value 0.15, and 0 and 1 have node-types C and Y, respectively, this file should contain a line of 0\t1\t0.15\tCY, where \t is TAB.

query.txt is a file that specifies query nodes included in data_graph.txt. Specifically, if we select node-ids 0 and 1 as query nodes, query.txt should be as follows:

0
1

SchemaRank reads the given files by using the CRS format, and this thus requires a file conversion process. To covert the input files (schema_graph.txt and data_graph.txt) into the CRS format (output_sg.bin and output_dg.bin), we need to run gen_crs like as follows:

$ cd ./generate_crs/
$ ./gen_crs -s schema_graph.txt -d data_graph.txt

gen_crs requires two options, -s and -d, that specify schema_graph.txt and data_graph.txt, respectively.

Finally, we can run SchemaRank like as follows:

$ ./schemarank -i <sg_crs_file> <dg_crs_file> -q query.txt 

where <sg_crs_file> and <dg_crs_file> are CRS files of a schema and a data graph obtained by generate_crs, respectively.

To learn further options of schemarank, please run ./schemarank -h.

• http://www.kde.cs.tsukuba.ac.jp/~shion/index.html.en