Writing on Clouds Vadim Mazalov and Stephen M. Watt Department of Computer Science The University of Western Ontario London Ontario, Canada N6A 5B7 {vmazalov,Stephen.Watt}@uwo.ca Abstract. While writer-independent handwriting recognition systems are now achieving good recognition rates, writer-dependent systems will always do better. We expect this difference in performance to be even larger for certain applications, such as mathematical handwriting recog- nition, with large symbol sets, symbols that are often poorly written, and no fixed dictionary. In the past, to use writer-dependent recogni- tion software, a writer would train the system on a particular computing device without too much inconvenience. Today, however, each user will typically have multiple devices used in different settings, or even simulta- neously. We present an architecture to share training data among devices and, as a side benefit, to collect writer corrections over time to improve personal writing recognition. This is done with the aid of a handwriting profile server to which various handwriting applications connect, refer- ence, and update. The user’s handwriting profile consists of a cloud of sample points, each representing one character in a functional basis. This provides compact storage on the server, rapid recognition on the client, and support for handwriting neatening. This work uses the word “cloud” in two senses. First, it is used in the sense of cloud storage for informa- tion to be shared across several devices. Secondly, it is used to mean clouds of handwriting sample points in the function space representing curve traces. We “write on clouds” in both these senses. Keywords: Handwriting Recognition, Mathematical Handwriting Recog- nition, Cloud Computing, Service Oriented Architecture 1 Introduction We are interested in online recognition of handwritten mathematics. The wide- spread use of hand-held mobile devices and tablets has created a ubiquitous environment for two-dimensional math input. Writing mathematics on a digital canvas is similar to traditional pen-on-paper input. It does not require learn- ing any typesetting languages and can be efficient, given a robust and reliable implementation. According to one study [1], pen-based input of mathematics is about three times faster and two times less error-prone than standard keyboard- and mouse-driven techniques. However, recognition of mathematics is a harder problem than recognition of natural language text.