首页|Patent Issued for Iteratively trained machine learning models for evaluations of internal consistency (USPTO 11977603)
Patent Issued for Iteratively trained machine learning models for evaluations of internal consistency (USPTO 11977603)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
The following quote was obtained by the news editors from the background informa tion supplied by the inventors: “Categorizing a large number of items can be tim e-consuming, tedious, and internally inconsistent. Frequently, assigning categor ies to items involves judgment calls which may be error prone or subject to shif ts in perspective. For example, if 10,000 tasks are to be categorized according to 15 task types, an incorrect or not preferred category may be assigned to a ta sk as a result of, for example, a misunderstanding or misinterpretation of the t ask or the available categories. Often, there may be inconsistencies in how item s are categorized if, for example, different persons are categorizing the tasks and each interprets items or categories slightly differently, if the same person views items slightly differently on different days, and/or if a person’s interp retation of tasks and categories evolves during the process. A person’s categori zation of his or her one thousandth item may be more informed, performed with a different degree of care, and/or approached differently from the person’s approa ch in categorizing his or her tenth or hundredth item, such that the same item m ight be categorized one way if encountered at one point in time, but differently if encountered by the same person at another point in time. Moreover, reviewing the quality of categorizations also tends to be time consuming and inconsistent . For example, different persons checking and rechecking prior categorizations a t different times and with different perspectives may catch some obvious errors or otherwise refine categories for certain items, but may do so inconsistently, as the review may also be performed by multiple people with varying and/or evolv ing perspectives and approaches over potentially many days, weeks, or months. So metimes, an original categorization of an item may be preferable over a recatego rization during quality review by the same person (but, e.g., on a different day ) or by another person (who may, e.g., have a different perspective or approach) .”
BusinessCyborgsEmerging TechnologiesMachine LearningSupervised LearningWells Fargo Bank N.A
NETL
NSTL
