ChemInk: a natural real-time recognition system for chemical drawings
Proceedings of the 16th international conference on Intelligent user interfaces
| Hi-index | 0.00 |
In chemistry, molecules are drawn on paper and chalkboards as diagrams consisting of lines, letters, and symbols which represent not only the atoms and bonds in the molecules but concisely encode cues to the 3D geometry of the molecules. Recent efforts into pen-based input methods for chemistry software have made progress at allowing chemists to input 2D diagrams of molecules into a computer simply by drawing them on a digitizer tablet. However, the task of interpreting these parsed sketches into proper 3D models has been largely unsolved due to the difficulty in making the models satisfy both the natural properties of molecule structure and the geometric cues made explicit in the drawing. This dissertation presents a set of techniques developed to solve this model construction problem within the context of an educational application for chemistry students. Our primary contribution is a framework for combining molecular structure knowledge and molecule diagram understanding via augmenting molecular mechanics equations to include drawing-based penalty terms. Additionally, we present an algorithm for generating molecule models from drawn diagrams which leverages domain-specific and diagram-driven heuristics. These heuristics make our process fast and accurate enough for molecule diagram drawing to be used as an interactive technique for model construction on modern Tablet PC computers.