Online structural analysis for handwritten chemical expression recognition

With recent emergence of pen-based and touch-based input surfaces, it now becomes more feasible to input chemical expressions directly by handwriting, which is in fact more natural and efficient. In general, chemical expressions can be written in either molecular or structural format. However, it is typically more difficult and challenging to directly take structural chemical expressions as input because we need to further deal with complex two-dimensional structures in the handwriting. In this paper, we propose an effective and novel structural analysis approach to support online handwritten chemical expression recognition. Unlike off-line handwriting recognition, our approach can analyze the chemical structure progressively and present recognition results to the user immediately after each chemical symbol is drawn. In detail, we develop a hybrid SVM and Elastic Matching approach together with a user feedback mechanism to enable efficient and structural recognition of chemical expressions. As presented in the result shown later in the paper, our system is able to achieve very good performance with high accuracy.