FIG. 20 is a flowchart of a process 2000 for identifying a handwriting language type. In some aspects one or more of the functions discussed below with respect to FIG. 18 may be performed by hardware processing circuitry. For example, instructions 1124 stored in a memory, such as one or more hardware memories (e.g. 1104 and/or 1106), configure the hardware processor 1102 to perform one or more of the functions discussed below with respect to FIG. 20.

In operation 2010, a length and height of a region of an image is determined. The image may be of a document. The region may be a bounding box defined to substantially surround an area of an image having a probability of including handwriting. As discussed above with respect to operation 1705, some embodiments may analyze an image of a document and identify one or more regions in the image having a handwriting probability above a threshold. The length and height of the region are determined in operation 2010. In some aspects, operation 2010 determines the height of the region to be parallel to a direction of a majority of linear features in the region. For example, operation 2010 may analyze features within the region and identify linearity to groups of features in the region. Option 2010 may then determine a vector that a majority of the groups of features are aligned with. The height may be substantially parallel to this vector while the length may be substantially perpendicular with the vector.