Gastrointestinal infections (GI) affect humans and animals. The symptoms of GI include mild fever (up to 39 ℃ temperature in humans) and vomiting. In the UK, the number of confirmed human laboratory reports for GI has increased from 2023 to 2025. In early 2020, there were ongoing concerns from veterinary practitioners and owners about a UK outbreak of GI affecting dogs (around 38 °C–39 ℃ is a dog's normal temperature). This paper investigates symbolic rule-based concept extraction for temperature and vomiting (useful for syndromic surveillance) from veterinary clinical narratives. To validate the symbolic approach proposed to build concept detectors, we used more than 1 million consults collected over 12 months from UK veterinary practices that supported the UK outbreak investigation of GI in dogs. The concept detectors we built leverage on traditional methods from symbolic Artificial intelligence (AI), exploiting domain specific knowledge. A state-of-the-art alternative to the concept detectors is Large Language Models (LLMs) from neural AI. However, LLMs have raised ethical concerns, such as data ownership and explainability. We explored the benefits of using the concept detectors to: (a) annotate datasets for customising (fine-tuning) open-source LLMs to detect mentions of temperature and vomiting; and (b) provide outcome explanations for LLM’s predictions.