Word bank-based OCR semantic correction method and system, medium, equipment and terminal

Abstract

The invention discloses a word bank-based OCR semantic correction method and system, a medium, equipment and a terminal, and relates to the technical field of semantic networks. The method comprises: filtering out all the characters of which the character recognition confidence is smaller than a threshold value 0.4, and recording a result of the first five of each character; extracting suspicion words from the five results of the characters with the confidence coefficient smaller than a threshold value 0.4, and finally, correspondingly obtaining five groups of suspicion words, wherein each group of suspicion words can have one or more suspicion words; respectively matching the five groups of suspicion words and the word bank according to the sequence of confidence from high to low, respectively calculating matching distances, and evaluating the matching distances by adopting an editing distance; and outputting the result with the minimum matching distance as a result. The character recognition result top5 and the word bank are combined, the wrong characters are accurately positioned and recognized, efficient error correction is performed on the recognized wrong characters in a targeted manner, and the semantic correctness of the recognition result is ensured.

Description

technical field [0001] The invention belongs to the technical field of semantic network, and in particular relates to a thesaurus-based OCR semantic correction method, system, program storage medium for receiving user input, computer equipment, and information data processing terminal, which can be applied to artificial intelligence, deep learning, and computer image vision . Background technique [0002] OCR (Optical Character Recognition) technology has basically matured in the recognition of texts with better printing quality, and some software on the market can achieve a very high single-word recognition rate for these texts. However, for some texts with low printing quality or ambiguous handwriting, the single-character recognition rate of OCR will drop significantly, and post-processing technology must be relied on to improve the overall recognition rate of the full text. At present, there are two commonly used and effective methods in post-processing technology. One ...

AI Technical Summary

Problems solved by technology

[0008] (1) The common word matching technology needs to match all the vocabulary of the recognized sentence, and the efficiency of blind error correction without targeting is low;

[0009] (2) The semantic correction method based on natural language processing technology is suitable for detecting wrong words, but it is difficult to predict correct words and prone to miscorrection;

[0010] (3) The semantic correction method based on natural language processing technology requires a large number of annotations and predictions, which is difficult to implement

Images