Methods about Deep Learning for Text Matching
Understanding the Methods in Text Matching Area Including Key-words based Matching Model & Latent Semantic Matching Model. Implement the Classical Methods.
DSSM
Learning Deep Structured Semantic Models for Web Search using Clickthrough Data
CIKM 2013
词袋模型,基于语义表达的结构, word hash + DNN
详细解释
代码
CDSSM
Learning Semantic Representations Using Convolutional Neural Networks for Web Search
WWW 2014, word hash + CNN + DNN
CLSM
A Latent Semantic Model with Convolutional-Pooling Structure for Information Retrieval
CIKM 2014
基于匹配的结构, word hash + CNN, CLSM和C-DSSM有什么区别呢
DSSM的应用
Modeling Interestingness with Deep Neural Networks
EMNLP 2014
DSSM应用于文本分析,在automatic highlighting和contextual entity search问题上效果好。
主要有两点贡献:
1) DSSM + CNN
2) 不针对相关性,加了一个ranker
ARC-I/ARC-II
Convolutional Neural Network Architectures for Matching Natural Language Sentences
NIPS 2014
CNN的基于语义表达和基于匹配的两种结构; 增加了门解决句子长度不一致问题
CNTN
Convolutional Neural Tensor Network Architecture for Community-based Question Answering
IJCAI 2015
(D)CNN+MLP(tensor layer);
基于语义表达的结构
DeepMatch
A Deep Architecture for Matching Short Texts
NIPS 2013
Reviews
目的:建模更复杂的匹配关系。最早的基于匹配的结构把。
结合了localness和hierarchy intrinsic,基于点积的网络不好做的,最大的亮点是用话题模型建立网络吧。
DSSM/Sent2Vec Release Version
MSRA发布的Sent2Vec发行版
https://github.com/robertsdionne/neural-network-papers/blob/master/README.md