Analytics platforms such as IBM's Watson AnalyticsTM are collecting metadata about their use, including user queries on uploaded datasets. The analysis of this metadata may be valuable in improving services, such as query recommendation and automatic data visualization. However, analysis of metadata is difficult not only in terms of scale but also in terms of complexity. Generalizing and exploring query patterns across users and datasets is challenging. Abstractions are likely to help bridge differences in specifics (e.g., column names and query details), particularly in semantics. For example, a single query, "What is the trend of sales over year?" could be abstracted in many different ways (e.g., "What is the trend of financial gain over time?"). In this paper, we describe our process of creating a dataset of query semantics, starting from initial metadata extraction from query logs to semantic expansion using WordNet. To help system designers effectively browse and understand patterns of use, we developed VIQS (Visual Interactive Query Semantics), a system that extracts query semantics from query logs over multiple datasets, and allows users to explore underlying patterns visually. We present results from an informal interview study along with specific insights regarding popular query patterns from 3-months of data. We believe the analytic process, as well as the specific insights on query patterns, will benefit the design of analytics platforms. CCS Concepts •Human-centered computing ! Systems and tools for interaction design; Information visualization;.