A Neuro-Symbolic System over Knowledge Graphs for Link Prediction

Tracking #: 3324-4538

Ariam Rivas
Diego Collarana
Maria Torrente
Maria-Esther Vidal

Responsible editor: 
Guest Editors NeSy 2022

Submission type: 
Full Paper
Neuro-Symbolic Artificial Intelligence (AI) focuses on integrating symbolic and sub-symbolic systems to enhance the performance and explainability of predictive models. Symbolic and sub-symbolic approaches differ fundamentally in how they represent data and make use of data features to reach conclusions. Neuro-symbolic systems have recently received significant attention in the scientific community. However, despite efforts in neural-symbolic integration, symbolic processing can still be better exploited, mainly when these hybrid approaches are defined on top of knowledge graphs. This work is built on the statement that knowledge graphs can naturally represent the convergence between data and their contextual meaning (i.e., knowledge). We propose a hybrid system that resorts to symbolic reasoning, expressed as a deductive database, to augment the contextual meaning of entities in a knowledge graph, thus, improving the performance of link prediction implemented using knowledge graph embedding (KGE) models. An entity context is defined as the ego network of the entity in a knowledge graph. Given a link prediction task, the proposed approach deduces new RDF triples in the ego networks of the entities that correspond to the heads and tails of the prediction task on the knowledge graph (KG). Since knowledge graphs may be incomplete and sparse, the facts deduced by the symbolic system not only reduce sparsity but also make explicit meaningful relations among the entities that compose an entity ego network. As a proof of concept, our approach is applied over a KG for lung cancer to predict treatment effectiveness. The empirical results put the deduction power of deductive databases into perspective. They indicate that making explicit deduced relationships in the ego networks empowers all the studied KGE models to generate more accurate links.
Full PDF Version: 


Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 03/Mar/2023
Review Comment:

Authors have revised the paper which is now much more readable and understandable the before.
The paper is much more clear and more easy to follow. Examples have also been improved and are helping a lot to follow the points of the authors.
The paper illustrates well what could be a neuro-symbolic system and what could mean neuro-symbolic integration. The application is also quite useful and well adapted to the purpose.
Moreover, the presentation of examples has been improved and it is much more easy to read and understand the curves and the results in pages 18 and 19.
The paper can be accepted in the journal.
A few comments are following and should be checked by the authors.

In page 2, line 16, "ego network" is mentioned but not explained.

In page 3, line 23, DS is used but this abbreviation should also be made explicit, such as DS (deductive system).

In page 4, line 25, the sentence is strange, should it be instead: Note that all relationships in KG_comp are NOT necessarily true?

Again writing "true relationships" is a bit misleading, maybe authors should complete as "relationships declared as true (w.r.t. domain knowledge for example)".

in page 5,
in line 9: KGE is a knowledge graph embedding model over KG...
in line 41: KGE is a machine learning model...
Should we consider that this is the same thing? In this case authors should explain why.

line 11: a Horn clause (and not clauseS)

lines 16 and 17: what is a "limited variable"?