Knowledge graphs (KGs) are a powerful paradigm that produces structured representations of heterogeneous data for semantic data integration. Nevertheless, it remains a challenge to ensure rich and up-to-date KGs. Recently, large language models (LLMs) have demonstrated capabilities in understanding language, offering opportunities to automate KG construction and enrichment. In this work, we address the problem of enriching KGs with contextual information collected from multiple sources. We propose a context-aware hybrid neuro-symbolic approach. This approach combines the capabilities of language understanding of LLMs with the formal reasoning which is provided by semantic representations. Our approach supports contextual triple generation, semantic alignment of entities and relations, constraint validation, and dynamic ontology evolution. We conduct experiments on real-world KG and compare the proposed approach to state-of-the-art methods. Moreover, it provides a transparent process by involving users in the loop and ensures a more trustworthy human–AI collaboration.