Introduction

---

1. Term Frequency - Inverse Document Frequency
2. t-distributed Stochastic Neighbor Embedding
3. Article Spinning
4. Text classification -- sentence embedding average embeddings of all words + naive classifier/logistic regression
5. Bigram model -- probability of transition occurence / softmax logistic regression / 1 hidden layer neural network
6. Word2Vec Embedding -- Continuous Bag Of Words (Many to one), Skip-Gram (One to many), Hierachical Softmax (like Huffman Code), Negative Sampling (like Triplet Loss)
7. GloVe Embedding -- like Recommendation System
8. POS NER -- HMM, DNN with Embedding and tree embedded RNN
9. Bidirectional RNN -- Text generation
10. Seq2seq -- Encoder + Decoder Structure + Teacher Forcing (feedback right word even if the network has wrong previous word prediction)
11. Attention -- Encoder + Context + Decoder
12. Memory Networks -- train to make question embedding close to answer embedding + 2 Supporting Fact (pass first fact first, next next)

References

---