Extension of the word2vec concept
The word2vec concept can be extended to different levels of text. This concept can be applied on the paragraph level or on the document level, and apart from this, you can also generate the global vector, which is called GloVe. We will try to understand them. Here, we are going to get an overview of each of the concepts.
Here are the following extended concepts built by using the word2vec concept:
- Para2vec
- Doc2vec
- GloVe
Para2Vec
Para2vec stands for paragraph vector. The paragraph vector is an unsupervised algorithm that uses fixed-length feature representation. It derives this feature representation from variable-length pieces of texts such as sentences, paragraphs, and documents.
Para2vec can be derived by using the neural network. Most of the aspects are the same as Word2vec. Usually, three context words are considered and fed into the neural network. The neural network then tries to predict the fourth context word. Here, we are trying to maximize the log probability...
