huajiechen / att-cbow

Attention Model for Continuous Bag of Words

Requiement

• Python2.7
• Tensorflow0.12

Data

Data format is the same with gensim.Word2Vec. Each line is an document. And the text should be segmente in advance. Data in my experiment can be download from my homepage.

Parameters

• filename: location of input fille
• batch_size: size of batch
• num_epochs: train times
• window_size: context window size
• min_time: minimum times of word
• embedding_size: embedding size of word and character
• num_sampled: negtive sampled size
• start_lr:learning rate