Smart object description to metadata validator for online trading
Publication Date: 2017-Aug-03
The IP.com Prior Art Database
ABSTRACT: Items offered in online trading services come with metadata (used for auction
filtering, etc) and more verbose description including photos of the object, etc . For example
during online auctions the only descriptions of product are photos and description. The more
valuable the product is, the more important is that description will be valid. If the vase which is
said to be from XIII century has for example ornaments from XIX century, or old car has
headlights covers not matching its original design, it should be documented in description, or
else should be recognized and reported to user.
Proposed is a system that would use its embedded expert system to validate provided materials
vs provided description and possibly as well find some suspicious inconsistencies between
description/multimedia/metadata and so on.
Main idea is to
- process the description to detect named entities
- compare the named entities vs auction metadata
- process all the attached multimedia to assess its relevance to the above
- extract non-matching/suspicious elements to generate a report
E.g.: Catch meaningful phrases from description which are describing the object. Then check photos of
object looking for inconsistent elements regarding documented age of object. At the end description is
compared with prepared description of object using photos and all inconsistencies are reported.
1. Problem domain research Regarding on domain, info about problem is gathered.
For example: If the product is an old car the information from car forums is gathered regarding: * the most popular car modifications domain_info.modifications = ["changed windshield"] * the earliest dates of production car and its elements domain_info.elements = overall: 1990, headlights: 1989
* the most important elements of this car model domain_info.important_elements = [windshield, headlights, wheels] * the most characteristic elements of this car model domain_info.characteristic_...