We have divided the hundreds of thousands of registered items into several classes according to the purpose of use. Within each class, we have created and trained a model with intelligent intelligence to check and sort the items, which will check the items and detect duplicate records even though their names are not exactly identical. The output of such a check is a list of duplicates for each class of items, from which the client can easily identify whether the item is intentionally duplicated, or whether the duplication needs to be corrected.