The alignment class contains all the information available associated with an alignment across an ensemble of topic models. The available accessor methods are,

Details

* weights: Extract weights between all pairs of topics within an alignment. Topic pairs with high alignment scores are more similar to one another, though the precise implementation will depend on the method used during align_topics. Note that only the weights are needed in order to compute stability, refinement, and key topics summaries. * models: Extract the model parameters that were used in the original alignment. Note that the latent topics may have been reordered, to maximize the consistency across all models according to their alignment. * n_topics: How many topics are there total, within the alignment object? n_models: How many models total are there, within the alignment object?

See also

align_topics