介绍基于传统机器学习与基于前沿深度学习的语言模型的原理、实现以及应用。介绍词表示向量的具体概念,获取方式。在此基础上,详细分析讲解如BERT,GPT等前沿大语言模型。内容包括: (1) Traditional Language Model; (2) Smoothing; (3) Distributed Representation; (4) Word2Vector Algorithm; (5) Language Model: Training; (6) Language Model: Evaluation; (7) Neural Network Language Models; (8) Continuous bag-of-words (CBOW); (9) Skip-gram Algorithm; (10) Hierarchical softmax & Negative sampling (11) Glove Algorithm; (12) Recurrent Neural Network-based Language Models; (13) Seq2seq Model; (14) Attention Mechanisms; (15) Self-attention; (16) Transformer; (17) Pre-training a Language Model; (18) GPT: Generative Pretrained Transformer; (19) BERT: Bidirectional Encoder Representations from Tranformers. |