Store Devices Microsoft Surface PCs & tablets Xbox Virtual reality Accessories Windows phone Microsoft Band Software Office Windows Additional software Apps All apps Windows apps Windows phone apps Games Xbox One games Xbox 360 games PC games Windows games Windows phone games Entertainment All Entertainment Movies & TV Music Business & Education Business Store Education Store Developer Sale Back-to-school essentials Sale Products Software & services Windows Office Free downloads & security Internet Explorer Microsoft Edge Skype OneNote OneDrive Microsoft Health MSN Bing Microsoft Groove Microsoft Movies & TV Devices & Xbox All Microsoft devices Microsoft Surface All Windows PCs & tablets PC accessories Xbox & games Microsoft Band Microsoft Lumia All Windows phones Microsoft HoloLens For business Cloud Platform Microsoft Azure Microsoft Dynamics Windows for business Office for business Skype for business Surface for business Enterprise solutions Small business solutions Find a solutions provider Volume Licensing For developers & IT pros Develop Windows apps Microsoft Azure MSDN TechNet Visual Studio For students & educators Office for students OneNote in classroom Shop PCs & tablets perfect for students Microsoft in Education Support Sign in Research Research o Research Home o Research areas Algorithms Artificial intelligence and machine learning Computer systems and networking Computer vision Data visualization, analytics, and platform Ecology and environment Economics Graphics and multimedia Hardware, devices, and quantum computing Human-centered computing Mathematics o o o o o Medical, health, and genomics Natural language processing and speech Programming languages and software engineering Search and information retrieval Security, privacy, and cryptography Social Sciences Technology for emerging markets Products & Downloads Programs & Events Academic Programs Events & Conferences People Careers About About Microsoft Research blog Asia Lab Cambridge Lab India Lab New England Lab New York City Lab Redmond Lab Applied Sciences Lab Research areas o Algorithms o Artificial intelligence and machine learning o Computer systems and networking o Computer vision o Data visualization, analytics, and platform o Ecology and environment o Economics o Graphics and multimedia o Hardware, devices, and quantum computing o Human-centered computing o Mathematics o Medical, health, and genomics o Natural language processing and speech o Programming languages and software engineering o Search and information retrieval o Security, privacy, and cryptography o Social Sciences o Technology for emerging markets Products & Downloads Programs & Events o Academic Programs o Events & Conferences People Careers About o About o Microsoft Research blog o Asia Lab o Cambridge Lab o India Lab o New England Lab o New York City Lab o Redmond Lab o Applied Sciences Lab A Supervised Learning Approach to Search of Definitions April 1, 2006 Download Document BibTex Authors Jun Xu Yunbo Cao Hang Li Min Zhao Yalou Huang Publication Type Inproceedings Pages 19 Number MSR-TR-2006-18 Abstract Related Info Abstract This paper addresses the issue of search of definitions. Specifically, given a term, we are to find definition candidates of the term and rank the candidates according to their likelihood of being good definitions. This is in contrast to the traditional approaches of either generating a single combined definition or outputting all retrieved definitions. Necessity of conducting the task in practice is pointed out. Definition ranking is essential for the task. A specification for judging the goodness of a definition is given. In the specification, a definition is categorized into one of the three levels: ‘good definition’, ‘indifferent defi-nition’, or ‘bad definition’. Methods for performing definition ranking are also proposed in this paper, which formalize the problem as either classification or ordinal regression. We employ SVM (Support Vector Machines) as the classification model and Ranking SVM as the ordinal regression model respec-tively, such that they rank definition candidates according to their likelihood of being good definitions. Features for constructing the SVM and Ranking SVM models are defined, which represent the characteristics of term, definition candidate, and their relationship. Experimental results indicate that the use of SVM and Ranking SVM can significantly outperform the baseline methods of using heuristic rules, em-ploying the conventional information retrieval method of Okapi, or using SVM regression. This is true both when the answers are paragraphs and when they are sentences. Experimental results also show that SVM or Ranking SVM models trained in one domain can be adapted to another domain, indicating that generic models for definition ranking can be constructed. Related Info Research Areas Data visualization, analytics, and platform Research Labs Microsoft Research Lab - Redmond Follow Microsoft Research Follow @MSFTResearch Share this page Tweet Learn Windows Office Skype Outlook OneDrive MSN Devices Microsoft Surface Xbox PC and laptops Microsoft Lumia Microsoft Band Microsoft HoloLens Microsoft Store View account Order tracking Retail store locations Returns Sales & support Downloads Download Center Windows downloads Windows 10 Apps Office Apps Microsoft Lumia Apps Internet Explorer Values Diversity and inclusion Accessibility Environment Microsoft Philanthropies Corporate Social Responsibility Privacy at Microsoft Company Careers About Microsoft Company news Investors Research Site map English (United States) Contact us Privacy & cookies Terms of use Trademarks About our ads © 2016 Microsoft ​