eGitty

Discover The Most Popular Algorithms

Learning Personalized Word Representations with Negative Sampling

When we use word2vec or glove to learning word representations, word representations are able to capture syntactic and semantic regularities in text. However, it can not capture personalized information, such as sex, age or identification.

Paper: Personalized Semantic Word Vectors used authorship information to create personalized word representations.

We have knonw use document information to create document embeddings using negative sampling method.

Creating Document Embeddings with Negative Sampling for Document-level Sentiment Classification

If we replace document information with authorship, we also can create personalized word representations with negative sampling method.

How about personalized word representations?

As to movie review dataset, personalized word representations have been used for sentiment analysis. The comparative result as follow:

How about personalized word representations?

Leave a Reply

Your email address will not be published. Required fields are marked *