Vector space models (VSMs) of semantics are a promising approach for representing and processing the meaning of human language. This paper surveys the use of VSMs for semantic processing of text, organizing the literature according to the structure of the matrix in a VSM. Three broad classes of VSMs are identified: term–document, word–context, and pair–pattern matrices, each with corresponding applications. The paper provides a detailed look at specific open source projects in each category and highlights the breadth of applications of VSMs for semantics. It also offers a new perspective on VSMs for those already familiar with the area and provides pointers into the literature for those less familiar with the field. VSMs are based on the distributional hypothesis, which posits that words occurring in similar contexts tend to have similar meanings. This hypothesis is closely related to VSMs, which use event frequencies to derive vector representations of words, phrases, and documents. VSMs have been successfully applied to tasks such as measuring the similarity of meaning between words, phrases, and documents, and are used in search engines and other natural language processing applications. The paper discusses the motivation for VSMs, their relation to the distributional hypothesis, and their use in AI and cognitive science. It also highlights the importance of event frequencies in VSMs and the differences between various types of similarity. The paper provides a framework for organizing the literature on VSMs, including term–document, word–context, and pair–pattern matrices, and discusses the steps involved in generating a matrix, including linguistic and mathematical processing. The paper also discusses the use of VSMs in various applications, including information retrieval, semantic similarity, and semantic relations. It highlights the success of VSMs in these areas and their potential for further research. The paper concludes with a discussion of alternatives to VSMs for semantics and the future of VSMs, raising questions about their power and limitations.Vector space models (VSMs) of semantics are a promising approach for representing and processing the meaning of human language. This paper surveys the use of VSMs for semantic processing of text, organizing the literature according to the structure of the matrix in a VSM. Three broad classes of VSMs are identified: term–document, word–context, and pair–pattern matrices, each with corresponding applications. The paper provides a detailed look at specific open source projects in each category and highlights the breadth of applications of VSMs for semantics. It also offers a new perspective on VSMs for those already familiar with the area and provides pointers into the literature for those less familiar with the field. VSMs are based on the distributional hypothesis, which posits that words occurring in similar contexts tend to have similar meanings. This hypothesis is closely related to VSMs, which use event frequencies to derive vector representations of words, phrases, and documents. VSMs have been successfully applied to tasks such as measuring the similarity of meaning between words, phrases, and documents, and are used in search engines and other natural language processing applications. The paper discusses the motivation for VSMs, their relation to the distributional hypothesis, and their use in AI and cognitive science. It also highlights the importance of event frequencies in VSMs and the differences between various types of similarity. The paper provides a framework for organizing the literature on VSMs, including term–document, word–context, and pair–pattern matrices, and discusses the steps involved in generating a matrix, including linguistic and mathematical processing. The paper also discusses the use of VSMs in various applications, including information retrieval, semantic similarity, and semantic relations. It highlights the success of VSMs in these areas and their potential for further research. The paper concludes with a discussion of alternatives to VSMs for semantics and the future of VSMs, raising questions about their power and limitations.