Halal or Not: Knowledge Graph Completion for Predicting Cultural Appropriateness of Daily Products

TL;DR

HaCKG addresses the challenge of predicting halal status for cosmetics by moving beyond isolated ingredient checks to modeling high-order relations via a cosmetic knowledge graph. It introduces a pre-trained residual relational Graph Attention Network that fuses numerical ingredient properties with entity embeddings, trained in a self-supervised manner and fine-tuned for halal-status prediction. The authors construct a large knowledge graph with 11 entity types and 5 relation types (over 100k entities), design a gate-based feature fusion, and validate the approach against strong baselines, achieving state-of-the-art results on halal prediction tasks. The work demonstrates that relational graph learning over a heterogeneous cosmetic knowledge graph can robustly infer cultural appropriateness and offers a scalable mechanism for halal certification support.

Abstract

The growing demand for halal cosmetic products has exposed significant challenges, especially in Muslim-majority countries. Recently, various machine learning-based strategies, e.g., image-based methods, have shown remarkable success in predicting the halal status of cosmetics. However, these methods mainly focus on analyzing the discrete and specific ingredients within separate cosmetics, which ignore the high-order and complex relations between cosmetics and ingredients. To address this problem, we propose a halal cosmetic recommendation framework, namely HaCKG, that leverages a knowledge graph of cosmetics and their ingredients to explicitly model and capture the relationships between cosmetics and their components. By representing cosmetics and ingredients as entities within the knowledge graph, HaCKG effectively learns the high-order and complex relations between entities, offering a robust method for predicting halal status. Specifically, we first construct a cosmetic knowledge graph representing the relations between various cosmetics, ingredients, and their properties. We then propose a pre-trained relational graph attention network model with residual connections to learn the structural relation between entities in the knowledge graph. The pre-trained model is then fine-tuned on downstream cosmetic data to predict halal status. Extensive experiments on the cosmetic dataset over halal prediction tasks demonstrate the superiority of our model over state-of-the-art baselines.

Figures (4)

• Figure 1: The overall architecture of HaCKG. The model comprises four main blocks: cosmetic knowledge graph construction, gate networks, relational Graph Attention Networks, and Optimization.
• Figure 2: An example of the product prediction given a knowledge graph. The pre-trained model can predict a score between a new product $P_3$ and the status $S_0$ or $S_1$.
• Figure 3: An example of the cosmetic knowledge graph. The circles and narrows denote entities and their relations, respectively. There are several entities, such as $P$, $B$, and $I$, denote the cosmetic product, cosmetic brand, and ingredients, respectively. The brackets $[\cdot]$ refer to the numeric attributes of the ingredient properties.
• Figure 4: The model accuracy according to the number of layers.