US 12,073,416 B2
Method using artificial neural networks to find a unique harmonized system code from given texts and system for implementing the same
Asim Barlin, Istanbul (TR)
Assigned to Solmaz Gumruk Musavirligi A.S., Sisli/Istanbul (TR)
Appl. No. 17/257,536
Filed by Solmaz Gumruk Musavirligi A.S., Istanbul (TR)
PCT Filed Jul. 4, 2018, PCT No. PCT/TR2018/050344
§ 371(c)(1), (2) Date Dec. 31, 2020,
PCT Pub. No. WO2020/009670, PCT Pub. Date Jan. 9, 2020.
Prior Publication US 2021/0312470 A1, Oct. 7, 2021
Int. Cl. G06Q 30/018 (2023.01); G06F 16/33 (2019.01); G06F 40/205 (2020.01); G06F 40/284 (2020.01); G06N 3/08 (2023.01); G06Q 10/087 (2023.01); G06Q 50/26 (2024.01)
CPC G06Q 30/018 (2013.01) [G06F 16/3347 (2019.01); G06F 40/205 (2020.01); G06F 40/284 (2020.01); G06N 3/08 (2013.01); G06Q 10/087 (2013.01); G06Q 50/26 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A method for allocating unique identification numbers to merchandise items, comprising steps of;
designing a plurality of artificial neural networks each of which includes a fully-connected hidden layer;
training the plurality of artificial neural networks with files in a form of training sets to produce a plurality of artificial neural network models;
forming a significance-based hierarchical artificial neural network sequence of the plurality of artificial neural network models organized as: one first-level neural network model for suggesting categories; one second-level neural network model for suggesting chapters; one third-level neural network model for suggesting positions; one fourth-level neural network model for suggesting subpositions, and one fifth-level neural network model for suggesting at least one identification number as a whole, wherein
an output of the first-level neural network model is used as an input of the second-level neural network model,
an output of the second-level neural network model is used as an input of the third-level neural network model,
an output of the third-level neural network model is used as an input of the fourth-level neural network model,
an output of the fourth-level neural network model is used as an input of the fifth-level neural network model, and
an identification number suggested by the fifth-level neural network model comprises a category suggested by the first-level neural network model, a chapter suggested by the second-level neural network model, a position suggested by the third-level neural network model, and a subposition suggested by the fourth-level neural network model;
introducing a preprocessed text vector regarding a merchandise item as input to the first-level neural network model of said artificial neural network sequence, and;
obtaining a unique identification number prediction to the merchandise item from said artificial neural network sequence.