Artifact Relationship Best Practices and Considerations

The SEARCCH hub supports the concept of artifact relationships in order to help users more easily find related artifacts (e.g., the software that was used to perform the research described in a paper.) Supported relationships are:

• cites
• extends
• supplements
• describes
• processes
• produces
• requires
• uses

Some relationships do not make sense between certain artifacts, so choosing the right relationship for each artifact pair is important. Table 1 shows good relationships between artifact types. The "uses" relationship is also provided in the event that one of the other relationships does not quite express the intended relationship.

Table 1. Best uses of supported artifact relationships 

artifact    >	        < (derived)	artifact     description

publication cites	cited by	publication  General citation of another publication

publication extends	extended by	publication  Work described builds on another prior work (stronger citation)

publication supplements supplemented by publication  Report provides supplemental material for another publication

publication describes	described by    software     Paper describes software implementation

publication describes	described by	dataset	     Paper describes dataset (can be input or output)

software    processes	processed by	dataset	     Software processes an input dataset

software    produces	produced by	dataset	     Software produces or creates an output dataset

software    requires	required by	software     Software requires another piece of software to work 

Here are some examples in practice:

• work described in a published conference paper (publication) extends work described in another published conference paper (publication)
• a dissertation (publication) provides supplemental information to a published conference paper (publication)
• a published conference paper (publication) describes an ML algorithm (software) and training data (dataset)
• an ML program (software) processes training data (dataset) to produce a trained model (dataset)
• an ML program (software) processes a trained model (dataset) and input data (dataset) to produce detections (dataset)
• wireshark (software) requires libpcap (software)
• a published conference paper (publication) describes an experiment (other)
• an experiment setup/configuration (software) requires a traffic generator (software) and an analysis program (software)

By specifying these kinds of relationships during the import process, you can help other researchers find all the artifacts associated with your body of research. 

When importing an artifact, you should specify the forward relationships. The hub will automatically derive and display the reverse relationships. 

If you need to add a forward relationship to an already published artifact that you own, please request using the "Send Us Feedback" form. 

If you need to add a forward relationship to an already published artifact that someone else owns, please contact the artifact owner.