Automatic query generation in guided systems Natural language generation from graphically built query Sara Paiva Viana do Castelo Polytechnic Institute Viana do Castelo, Portugal sara.paiva@estg.ipvc.pt Manuel Ramos-Cabrer Vigo University Vigo, Spain mramos@det.uvigo.es Alberto Gil-Solla Vigo University Vigo, Spain agil@det.uvigo.es Abstract— Guided-based systems are an alternative to natural language systems regarding the query construction method. We present a new solution that bridges the gap between these two types of systems by providing a hybrid interface which combines a graphical query interface for query construction with the correspondent natural language expression automatically generated by the system. Keywords: Natural language expressions, automatic generation, guided-based systems, query construction I. INTRODUCTION Semantic search has emerged as a way to increase the accuracy of the searching activity. The incorporation of semantics – meaning of the language – empowers search engines with a better understanding of the search expression which leads to more accurate results. Semantic search optimum vision is to allow the use of natural language (NL) input. However, this approach is still not able to produce accurate results as the query is not fully interpretated and understood by the system [1]. The guided construction method arises as an alternative for NL input. We refer to these systems as guided-based systems. They guide the user in the query construction what makes them less flexible than NL systems but they can produce totally accurate results what makes it a promising alternative. Despite this, there is still room for improvement in these systems. Guided-based systems use several approaches for building the query such as auto complete functionalities, menu-based or graphical interfaces. Our study is focused on the last option as we believe this query construction approach to be an advantage for the end user as it is simpler and intuitive. In this paper we introduce our solution which has a hybrid interface where the user can build the query graphically and, simultaneously, see it expressed in natural language. This is assured by the natural language generator engine which is the main focus of this paper. With this new functionality in guided-based systems, we intend to bring closer guided query construction methods to natural language ones and to make the user experience on building the query construction an easy process. II. GUIDED-BASED SYSTEMS INTERFACES As mentioned above, guided-systems use several approaches for building the query. Ginseng [2] uses an auto complete approach as the system predicts the possible completions of what the user entered in a free form entry field. Semantic Crystal [3], TAMBIS [4] and LingoLogic [5] all provide a graphical query construction method. The first two systems present much of the ontology´s formalisms (classes and properties are presented to the user with the same designation from the ontology which in the great majority is closest to a code than to a real word concept) which represents a disadvantage to the end user, not necessary familiar with this notation. LingoLogic’s interface is menu-based which makes it different from the two previous systems. The formalism issue raised above is well addressed in LingoLogic as the options shown to the user are real-world concepts. However, we find the menu-based approach less attractive and easier for the end user. Regarding the usability of the interfaces, Semantic Crystal presents the query as a graph which makes it hard to read and difficult to understand. TAMBIS has also an unfriendly query presentation as it relies on a presentation structure that resembles a tree, with a hierarchical form. As mentioned above, LingoLogic presents the user with the possible options in a menu. As the query is built, the options shown are already presented in order to make the query look like natural language. We see our solution as “the best of both worlds” as it allows the graphical creation of the query in order to facilitate this process to the user and, simultaneously, generate the correspondent natural language expression so the query is easily understood. III. OUR SOLUTION Our solution provides a hybrid interface as it offers a graphical interface for the query construction and a real-time natural language expression generated automatically by the system. This user interface proposal is based on conducted surveys [3] where users manifest their preference for a combination of two functionalities to construct the query: use of natural language and graphical construction. 2010 11th ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing 978-0-7695-4088-7/10 $26.00 © 2010 IEEE DOI 10.1109/SNPD.2010.33 165