Artemis - an extensible natural language framework for data querying and manipulation

Filtering and finding items of interest in large volumes of data, such as products in an e-commerce application or invoices in an ERP web platform can be a burdensome task, either for novice users that do not have insights on how the data is modeled or for those users who are already accustomed to the used system, but usually their filtering needs are significantly more complex. Natural language processing provides a way to dramatically improve the search experience for end-users and even though NLP is an AI-complete problem for the moment, based on the underlying data models the user can conduct comprehensive queries in a large set of scenarios guided by powerful context-aware prediction mechanisms. Based on this we have built Artemis, an extensible framework that transforms valid query inputs into filters applicable to underlying data models, stored either in-memory or in RDBMS, along with an extension mechanism to further enrich queries expressiveness via annotations or custom configuration and a context-aware prediction mechanism to direct the user into providing a valid query input, while decreasing the searching time. The conducted usability tests, showed that Artemis yields significantly reduced time-to-search, both for novice or experienced end-users with little or no training and for application developers it provides a straightforward apparatus for further improving expressivity based on custom, business-specific vocabulary.