Keyword recommendation method and system based on latent Dirichlet allocation model

摘要: Keyword recommendation methods and systems based on a latent Dirichlet allocation (LDA) model. The method comprises: calculating basic model for training texts; obtaining an incremental seed word, selecting from the texts text matching word to serve as text; probability distribution of complete words topics topics; relevance score between any other respectively obtain every two words; determining, according obtained query words, keyword corresponding word. By employing model, present invention greatly improves precision topic clustering diversity, significantly quality keywords in topics.