Artificial Intelligence for Big Data

Word2Vec

In a typical feature extraction from text, numerical vectors are created based on unique labels given to them. However, these uniquely-labeled sparse vectors do not represent the context in which each word has appeared. In other words, it does not specifically state or represent the relationship a given word exhibits with other words in a corpus. That means unsupervised learning algorithms that learn from data processing cannot be leveraged much. These algorithms cannot leverage relationships or contextual information about the word. Therefore, a new class of algorithms for feature extraction is developed that preserves the context or relationship information among words. This new class of algorithms is called Word-Embedding feature-extraction ...