Recommendation system
A system that selects or ranks relevant candidates from available items, for example using their features or interaction history.
How are recommendations produced?
A recommendation system selects candidates from available content or products and may rank them by expected relevance. It can use item features, known interests or past interactions. Recommendations may be personalised or relate to an item currently being viewed. The objective and eligible candidates must be defined: similarity or frequent joint use alone does not make a suggestion suitable for the specific need.
In a fictional tool shop, suitable accessories are to be suggested for the drill D-28. Frequently bought-together products may include accessories with an incompatible fitting. Selection therefore needs to be limited to the required compatibility using verified product attributes. The team plans checks with suitable and unsuitable accessories as well as incomplete attribute data. Without reliable evidence of compatibility, the presentation must not promise a confirmed fit. Assessment considers suitable suggestions and identified unsuitable ones; clicks alone do not show suitability.
Sources
Related terms
- Machine learningModels learn patterns from data and apply them to new inputs.
- Large language modelA large language model processes language using learned patterns. Fluent text proves neither factual accuracy nor access to current business data.
- Language model tokenTokens are processing units for language models. They do not reliably correspond to whole words and differ from design tokens.
- AI inferenceApplying an already trained model to inputs to obtain a result.
- LLM evaluationLLM evaluation tests model outputs against defined tasks and criteria. One good example does not establish overall quality.
- Model cardA model card documents an AI model’s purpose, foundations, evaluation conditions and known limitations for a particular version.
Articles on this topic
AI & Automation with ASES
AI assistants and automated workflows