Contextual Spelling Suggestions Based on Specialized Vocabulary and Statistical Information of Historical System Usage

Today¿s general purpose search engines have developed contextual based spelling/search suggestion software tools that are derived from the historical usage of the system. In other words, every search the system performs is stored and used as the basis of a statistical algorithm that locates contextual errors in a search. In the past several years, research analysts have identified a new tier of search called ¿specialized search,¿ which includes local, topical, and vertical searches. Vertical search engines, such as the DOE Office of Scientific and Technical Information¿s Information Bridge and Energy Citations Database, have not had the same success with the application of spelling/search suggestion due to the lower total population of searches performed against these sources. This project will develop a non-contextua,l pure spelling suggestion tool based on common spelling correction algorithms driven by word lists to be used with vertical search engines. Phase I will create a prototype spelling/search suggestion Web-based XML tool that combines historical usage and specialized vocabulary to contextually correct spelling or other search problems. Phase II will build upon the experimental data acquired from the prototype service, in order to develop a more robust Application Program Interface (API) tool and a more mature algorithm to create suggestions. Commercial Applications and other Benefits as described by the awardee: As vertical search engines gain popularity, the new tool should improve specialized search efficiency and effectiveness. Future benefits should include automation of historical data and specialized vocabulary building, thereby optimizing specialized searching for researchers, scientists, professionals, and academics.