US 11,880,660 B2
Interpreting text classifier results with affiliation and exemplification
Xinjian Xue, Bothell, WA (US); Adeel Jamil Siddiqui, Redmond, WA (US); Ryan Wang Gao, Seattle, WA (US); Jong-Chin Lin, Bellevue, WA (US); Bin Pan, Bellevue, WA (US); Naveen Duddi Haribabu, Redmond, WA (US); and Balaji Ramasubramaniam, Redmond, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Feb. 22, 2021, as Appl. No. 17/182,088.
Prior Publication US 2022/0269864 A1, Aug. 25, 2022
Int. Cl. G06F 40/30 (2020.01); G06F 16/35 (2019.01); G06F 40/279 (2020.01); G06N 3/044 (2023.01)
CPC G06F 40/30 (2020.01) [G06F 40/279 (2020.01); G06N 3/044 (2023.01)] 13 Claims
OG exemplary drawing
11. A computer-readable storage medium comprising instructions that, when executed by a processor, cause the processor to perform acts comprising:
obtaining an identity of a class previously assigned to an input vector by a computer-implemented text classifier, wherein the input vector is representative of a sequence of terms, and further wherein the class is amongst several classes to which input vectors are assignable by the computer-implemented text classifier;
based upon the identity of the class, selecting a class profile of a centroidal classifier, the class profile constructed based upon a set of vectors previously identified as belonging to the class by the computer-implemented text classifier; and
generating output based upon the selected class profile, wherein the output is configured to be information to a user of the computer-implemented text classifier as to why the computer-implemented text classifier assigned the input vector to the class.