Method, apparatus and electronic device for annotating information of structured document

A method for annotating information of a structured document, applied to an electronic device, comprising:

• performing a smearing treatment on an original image of the structured document according to a historical position of a field to be filled in the original image to obtain a template image of the structured document;
• obtaining at least one piece of annotation information of a field to be filled in the template image, wherein the annotation information comprises an attribute value of the field to be filled, a historical content of the field to be filled, and a historical position of the field to be filled in the template image;
• generating, according to the attribute value of the field to be filled, the historical content of the field to be filled and the historical position of the field to be filled in the template image, target filling information of the field to be filled, wherein the target filling information comprises:
• a target filling position, a target filling content and a target filling format;
• obtaining, according to the target filling information of the field to be filled, an image of an annotated structured document;
• and
• using the obtained image of the annotated structured document for training an image text recognition algorithm for extracting a structured document;
• wherein the generating, according to the attribute value of the field to be filled, the historical content of the field to be filled and the historical position of the field to be filled in the template image, the target filling information of the field to be filled, comprises:
• generating, according to the attribute value and the historical content of the field to be filled, the target filling content of the field to be filled;
• and
• performing a position adjustment according to the historical position of the field to be filled in the template image to obtain the target filling position of the field to be filled;
• wherein the generating, according to the attribute value and the historical content of the field to be filled, the target filling content of the field to be filled comprises:
• if the attribute value of the field to be filled exists in a preset corpus, randomly selecting a reference content from a content corresponding to the attribute value of the field to be filled in the preset corpus, and using the reference content as the target filling content of the field to be filled;
• if the attribute value of the field to be filled does not exist in the preset corpus and meanwhile a reference attribute value whose difference from the attribute value of the field to be filled is less than a preset value exists in the preset corpus, randomly selecting a reference content from a content corresponding to the reference attribute value in the preset corpus, and using the reference content as the target filling content of the field to be filled;
• and
• if neither the attribute value of the field to be filled nor the reference attribute value whose difference from the attribute value of the field to be filled is less than the preset value exist in the preset corpus, determining a part-of-speech of the historical content of the field to be filled, and generating the target filling content of the field to be filled according to whether a content corresponding to the part-of-speech of the historical content of the field to be filled exists in the preset corpus.