System and method for domain adaption with partial observation

摘要: System, method and computer program product provides a novel domain adaption/transfer learning approach applied to the problem of classifying abbreviated documents, e.g., short text messages, instant tweets. The proposed uses large number multi-labeled examples (source domain) improve on partial observations (target domain). Specifically, hidden, higher-level abstraction space is learned that meaningful for in source domain. This done by simultaneously minimizing document reconstruction error classification model hidden using known labels from target are then mapped same space, classified into label determined Exemplary results provided Twitter dataset demonstrate identifies topics useful classifications specific