Telecommunication networks pose challenges for root cause analysis due to complexity, heterogeneous data, and high event volumes. Existing models like RDF and LPG provide only partial solutions when used alone. This paper explores the Semantic-enhanced Programmable Graph (SPG) approach via the OpenSPG engine, which combines property graph flexibility with RDF-inspired semantics. Applied to a telecom use case, OpenSPG shows advantages for event-centric modeling but remains limited in maturity, tooling, and semantic interoperability, highlighting directions for future semantic-enabled network management.
To combine the strengths of RDF and property graphs, several mapping approaches have been developed.
G2GML [
This section outlines telecom domain challenges and introduces the OpenSPG engine as a hybrid solution for enterprise use cases.
There are several operational challenges faced by the network telecom systems2: Modern network operations face a multitude of data-related challenges that complicate efective monitoring, analysis, and automation. The overwhelming volume of data from various network planes (management, control, and data) combined with inconsistent storage formats and fragmented repositories, makes it dificult to analyse and extract actionable insights. Correlating this data across diferent sources, understanding service and customer impact, and preserving contextual information such as interface roles or link functions are particularly dificult. Additional complexities arise from varied data collection methods, organizational silos, and conflicting sources of truth (e.g., controllers vs. live network state). Altogether, these issues hinder the development of a unified, machine-readable, and semantically rich view of the network needed for autonomous decision-making and scalable operations.
Recent work has explored semantic solutions for network operations. Martínez et al. [
Released in late 2023, the OpenSPG engine provides an architecture for constructing domain knowledge graphs using a semantic schema, logical rules, and modular operators based on the SPG approach. It integrates LLMs for natural language processing and supports layered graph construction, reasoning, and querying. Data is stored in a graph database (default: Neo4j3), with schema definitions in MySQL; alignment between table columns and schema properties is essential for proper mapping. Column names must either be declared as schema properties or mapped to them programmatically. The OpenSPGSchema defines semantic predicates to enrich LPGs with syntax and semantic constraints. As shown in Fig. 1, it distinguishes three core types: Entity types (circles in Fig. 1) represent system objects (e.g., network elements); Event types (diamonds in Fig. 1) model actions afecting entities over time (e.g., route withdrawal); and Concept types (squares in Fig. 1) classify entities and events (e.g., control plane taxonomy). Distinguishing entity and event types supports semantic graph design by modeling entities as core objects and events as their interactions or state changes over time.
This section presents a network systems use case implemented with OpenSPG, illustrating schema modeling for causal event relationships. In a network environment, each router first receives advertised routes, then decides whether to accept them based on rules (like who sent them), and may pass them on to other peers. These interactions help routers build paths to reach diferent networks, enabling internet trafic to flow eficiently. A Border Gateway Protocol (BGP) router may lose connection to one of its upstream peers, triggering a route withdrawal. This means it no longer advertises certain network prefixes it previously learned from that peer. As a result, neighbouring routers must quickly ifnd alternative paths or may momentarily lose access to those networks. During this convergence period, packets destined for the withdrawn prefixes may be dropped, causing a temporary trafic outage or degradation in service until stable routes are re-established. Thus, it is imperative to identify the network elements which first causes the problem in the network.
Listing 1 presents a small script of the OpenSPG schema we created to represent relationships between diferent events. RouteWithdrawEvent is an event type which belongs to the event concept class TaxonomyControlPlane and it leads to the event type DroppedTrafficEvent. The subject of RouteWithdrawEvent is a network element. Designing a hybrid schema involves assigning labels to properties and defining schema constraints for classes. In the schema, it can be seen that NetworkElement has an index of on neID and hasNetworkElement property has a multivalue constraint.
‘TaxonomyControlPlane‘/‘RouteWithdraw‘: rule: [[
Define (e:RouteWithdrawEvent)-[p:belongTo]->(o:‘TaxonomyControlPlane‘/‘RouteWithdraw‘) {
Structure { } Constraint {
R1: e.index == ’route’
R2: e.trend == ’withdraw’
There are two ways of creating the relationships, one option is to introduce a new relation type in the schema which will require a manual upkeep of the schema and bring with it an associated risk of inconsistencies between the new data and the existing knowledge graph. A second approach is to create attributes and relationships through logical rules, with the goal of enabling more consistent and automated modeling. Listing 2 presents the rule for belongTo relationship where only index is route and trend is withdraw then the instances of RouteWithdrawEvent will belong to the concept of RouteWithdraw in the TaxonomyControlPlane. As a result of the rule given in Listing 2, we were able to classify diferent type of instances to their given classes dynamically. A Cypher query along with its corresponding result is shown in Fig. 2. For property graph management, OpenSPG relies on Neo4j as its underlying database.
This paper presents a work in progress investigation on the state-of-the-art methods, as part of their contribution to the ISWC industry track paper, reflecting both technical competence and a clear understanding of the underlying concepts in both OpenSPG and RDF approaches.
Switching from RDF-based modelling to the OpenSPG approach ofers flexibility with schema definitions but comes with notable trade-ofs in semantics and interoperability. Without RDF’s formal ontologies and reasoning capabilities, automated inference and standards-based validation (e.g., SHACL) are limited, increasing reliance on manual interpretation. Interoperability also sufers, as integration with external linked data sources, reuse of public RDF datasets, and portability between systems become more complex due to the lack of shared vocabularies and standard query protocols like SPARQL. While these drawbacks may be less critical in performance-driven, internal domains such as telecom event analysis, they reduce the universality and cross-system compatibility that are hallmarks of semantic web solutions. Our use case illustrates how causal relationships between network events can be presented by employing SPG for modeling complex, event-driven data in network operations. This early exploration highlights the potential of SPG-based systems in operational domains and lays the groundwork for future research in semantic modeling for large-scale, event-driven environments.
This research received funding from Research Ireland [grant number 24/IRDIF/13114] and ADAPT Researchers are partially funded by the European Regional Development Fund through the ADAPT Centre for Digital Content Technology [grant number 13/RC/2106].
During the preparation of this work, the authors used Chat-GPT-4 in order to: Grammar and spelling check. After using these tool(s)/service(s), the authors reviewed and edited the content as needed and take full responsibility for the publication’s content. [15] L. Tailhardat, Y. Chabot, R. Troncy, Noria-o: an ontology for anomaly detection and incident management in ict systems, in: European Semantic Web Conference, Springer, 2024, pp. 21–39.