Domain Heuristic Fusion of Multi-Word Embeddings for Nutrient Value Prediction

Being both a poison and cure for many lifestyle non-communicable diseases, food is inscribing itself into the prime focus of precise medicine. The monitoring few groups nutrients crucial some patients, methods easing their calculations are emerging. Our proposed machine learning pipeline deals with nutrient prediction based on learned vector representations short text–recipe names. In this study, we explored how results change when, instead using recipe description, use embeddings list ingredients. content one depends its ingredients; therefore, text ingredients contains more relevant information. We define domain-specific heuristic merging ingredients, which combines quantities each ingredient in order to them as features models prediction. from experiments indicate that improve when heuristic. protein were highly effective, accuracies up 97.98%. Implementing combining multi-word yields better than conventional heuristics, 60% accuracy cases.