Are All People Married? Determining Obligatory Attributes in Knowledge Bases

An attribute is obligatory for a class in a Knowledge Base (KB), if all instances of the class have the attribute in the real world. For example, hasBirthDate is an obligatory attribute for the class Person, while hasSpouse is not. In this paper, we propose a new way to model incompleteness in KBs. From this model, we derive a method to automatically determine obligatory attributes – using only the data from the KB. Our algorithm can detect such attributes with a precision of up to 90%.

Publication

Jonathan Lajus, Fabian M. Suchanek: “Are All People Married? Determining Obligatory Attributes in Knowledge Bases” (pdf)

Full paper at the Web Conference (WWW), 2018

Datasets

We ran our system on two datasets

YAGO (version 3.0.2)
Wikidata (Truthy 2017-06-27, see this)

Experimental Data

For each experiment, we report:

The results given by the different algorithms (here);
The gold standard used (here);
The per-instances evaluation for each methods and relations (here).

Raw results

In "data/Results/results" we provide tab-separated-value files of any class returned by any method, parameter and predicate.

IMPORTANT NOTE: here we only output upper-classes in the taxonomy

"predicate \t class" stands for "class(x) => \exists y. predicate(x, y)" and "predicate-1 \t class" stands for "class(y) => \exists x. predicate(x, y)"

Evaluation

We provide tab-separated-value files with the results of the experiments. They contain for every method, parameter, and predicate:

The number of true positives
The number of predicted instances
The number of instances derived from the gold standard
The precision, recall, and F1 per predicate
There is also their "NF" (for "New Facts") counterpart where we do not consider any instance already having the predicate in the KB (not used in the paper).

Source Code

The source code in the form of a Maven project can be found in the directory src

Dependencies not available in Maven are in src/lib

Runnables

All the jar files used for the experiments can be found in the run directory.

Refer to this documentation for usage information.

lajus/typing-www2018