j4nlp为java nlp项目,考虑到目前市面上主要是python及其生态库的天下,所以作者不打算使用开源深度学习库去实现,而是利用纯java和少量依赖包实现自然语言处理的一些常规模型,
项目结构,后期会更新和优化
- j4nlp
- common
- layers
- Affine
- Dropout
- Embedding
- MatMul
- Sigmoid
- SigmoidWithLoss
- Softmax
- SoftmaxWithLoss
- optimizer
- AdaGrad
- Adam
- Momentum
- Nesterov
- RMSprop
- SGD
- time_layers
- GRU
- LSTM
- RNN
- SimpleTimeAffine
- SimpleTimeSoftmaxWithLoss
- TimeAffine
- TimeBiLSTM
- TimeDropout
- TimeEmbedding
- TimeGRU
- TimeLSTM
- TimeRNN
- TimeSigmoidWithLoss
- TimeSoftmaxWithLoss
- trainer
- RnnLmTrainer
- Trainer
- BaseModel
- Functions
- Util
- layers
- core
- attention
- Attention
- AttentionDecoder
- AttentionEncoder
- AttentionSeq2Seq
- AttentionWeight
- TimeAttention
- WeightSum
- w2v
- negative_sampling_layer
- EmbeddingDot
- NegativeSamplingLoss
- UnigramSampler
- Cbow
- SkipGram
- negative_sampling_layer
- rnn
- SingleRnnLM
- MultiRnnLM
- seq2seq
- Decoder
- Encoder
- PeekyDecoder
- PeekySeq2Seq
- Seq2Seq
- attention
- dataset
- SeqDataset
- SpiralDataSet
- define_exception
- NumberParameterException
- utils
- common