| US 7,512,553 B2 | ||
| System for automated part-number mapping | ||
| Ghassan Chidiac, Wappingers Falls, N.Y. (US); Jayant R. Kalagnanam, Tarrytown, N.Y. (US); Moninder Singh, Farmington, Conn. (US); Sudhir Verma, Yorktown Heights, N.Y. (US); Fabio Dennis White, New Windsor, N.Y. (US); Michael D. Patek, Philadelphia, Pa. (US); and Yuk Wah Wong, Austin, Tex. (US) | ||
| Assigned to International Business Machines Corporation, Armonk, N.Y. (US) | ||
| Filed on Dec. 05, 2003, as Appl. No. 10/727,978. | ||
| Prior Publication US 2005/0125311 A1, Jun. 09, 2005 | ||
| Int. Cl. G06Q 10/00 (2006.01) | ||
| U.S. Cl. 705—28 [705/29; 705/22] | 1 Claim |
| 1. A system for automated mapping of part numbers associated with parts in bills of materials (BOMs), submitted by a plurality
of BOM originators to a BOM receiver, to the BOM receiver's internal part numbers, the system comprising a computer having
a processor, a memory operatively coupled to the processor, wherein the memory stores instructions that when executed by the
processor perform operations comprising:
receiving historical BOM data describing BOMs received by the BOM receiver, from a plurality of BOM originators, over a time
history;
receiving known mapping data defining historical mappings between the BOM receiver's internal assigned part numbers and the
BOM originators' various assigned part numbers;
receiving part description parameters describing a plurality of parts to which the BOM receiver has assigned internal part
numbers;
learning by computer methods mapping prediction models for predicting BOM internal part numbers based on received BOMs from
the plurality BOM originators, based on the historical BOM data, mapping data and part parametric data, wherein said computer
methods include feature extraction for tokenizing a textual description of parts according to a token scheme and generating
a corresponding list of parametric features based on the extracted tokens, wherein said computer methods form said learned
mapping prediction models according to a multi-level taxonomy arranged for a hierarchical prediction mapping, including initially
predicting a class of an unmapped part based on received information, and traversing down levels of the taxonomy, predicting
the sub-class of the unmapped part at each subsequent level and, at a leaf level of the multi-level taxonomy, classifying
the unmapped part to a BOM receiver internal part number based on the parametric features extracted from the BOM originator
textual description of the unmapped part;
learning said prediction models from said historical BOM data, known mapping data, and part description parameters, using
said methods of learning, and generating resulting learned prediction models;
receiving a BOM from a BOM originator, said BOM describing parts according to BOM originator assigned part numbers, wherein
at least one of said BOM originator assigned part numbers is an unmapped part number not within said historical mapping data;
predicting a BOM receiver internal part number associated with at least one part described by said BOM having an unmapped
part number, said predicting including applying at least one of said learned prediction models to said received BOM bill of
materials; and
generating a release data having said predicted BOM receiver internal part number.
|