# Curation: Pre-Annotation Before you spend time and budget on labeling, use FiftyOne to understand your raw data and choose *what* to annotate. Typical pre-annotation curation workflows include [visualizing embeddings](../brain/index.md#brain-embeddings-visualization) to explore the structure of your dataset, [similarity search](../user_guide/similarity.md#brain-similarity) (including [natural language queries](../user_guide/similarity.md#brain-similarity-text)), scoring samples by [uniqueness](../brain/index.md#brain-image-uniqueness), and pruning [near-duplicate](../brain/index.md#brain-near-duplicates) samples. In the FiftyOne App, the [Embeddings panel](../user_guide/app.md#app-embeddings-panel) and [similarity search](../user_guide/app.md#app-similarity-search-panel) make these workflows point-and-click, and [saved views](../user_guide/app.md#app-saving-views) let you turn the subsets you discover into named slices of your dataset.