Cosine similarity is a metric used to measure how similar two entities (usually in the form of vectors) are irrespective of their size. Mathematically, it calculates the cosine of the angle between two vectors projected in a multi-dimensional space. This approach is particularly useful in natural language processing and information retrieval where each term is considered a dimension, and a text document is represented as a vector in this high-dimensional space. By computing the cosine similarity, one can quantitatively estimate how similar two documents are likely to be in terms of their content. This is crucial for tasks like document similarity checks, clustering, and classification models.
The concept of cosine similarity originates from the field of vector algebra, where it has been used as a measure of the orientation of vectors to each other. Its adoption into computational fields like information retrieval and text mining became prevalent with the rise of vector space models in the 1960s and 1970s.
While the mathematical concept of cosine similarity is rooted in early mathematical studies of vectors, its application in computer science was heavily influenced by the development of the vector space model by Gerard Salton in the 1970s, particularly in the context of document retrieval systems. This adoption has since become foundational in fields dealing with large-scale text data.