Skip to main content

Edit this page

classification

The classification template runs a pre-trained text-classification model over the input and emits its predicted category labels as tokens. This lets you index documents by an inferred property — topic, language, sentiment — and then query them by label like any other term, without storing the label yourself.

How it works​

At index and query time the analyzer feeds the input text to a fastText supervised model loaded from modellocation and turns the model's predictions into terms. For example a cooking-topic model might turn "How do I bake a chocolate cake?" into the labels __label__cake, __label__chocolate and __label__baking. topk caps how many of the highest-scoring labels are emitted and threshold drops any whose confidence falls below the given score.

The model file is required and must be reachable from the server process at the path given in modellocation; the dictionary cannot be created without a loadable model. Classification and nearest_neighbors are the two model-backed templates — use classification to tag a document with predicted categories, and nearest_neighbors to expand it with semantically related terms.

Options​

OptionTypeDefaultDescription
modellocationstringrequiredPath to the fastText model file, reachable from the server
topkinteger1Number of top-scoring labels to emit
thresholddouble0.0Minimum confidence score to keep a label (0.0–1.0)

Usage​

Point modellocation at a trained fastText classifier and choose how many labels to keep:

Query
CREATE TEXT SEARCH DICTIONARY classifier (    template = 'classification',    MODELLOCATION = '/models/cooking.bin',    TOPK = 3,    THRESHOLD = 0.5);

With the dictionary attached to a text column in a USING inverted index, each document is indexed under its predicted labels, so a query for a label matches every document the model assigned it. Predicted labels are model-specific; the example above uses a small cooking-topic model and emits labels such as __label__cake.

See also​