Statistical relational learning for handwriting recognition

We introduce a novel application of handwriting recognition for Statistical Relational Learning. The proposed framework captures the intrinsic structure of handwriting by modeling fundamental character shape representations and their relationships using first-order logic. Our framework consists of three stages, (1) character extraction (2) feature generation and (3) class label prediction. In the character extraction stage, handwriting trajectory data is decoded into characters. Following this, character features (predicates) are defined across multiple levels - global, local and aggregated. Finally, a relational One-vs-All classifier is learned using relational functional gradient boosting (RFGB). We evaluate our approach on two datasets and demonstrate comparable accuracy to a well-established, meticulously engineered approach in the handwriting recognition paradigm.