17643224. INTENT CLASSIFICATION ENHANCEMENT THROUGH TRAINING DATA AUGMENTATION simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

From WikiPatents
Jump to navigation Jump to search

INTENT CLASSIFICATION ENHANCEMENT THROUGH TRAINING DATA AUGMENTATION

Organization Name

INTERNATIONAL BUSINESS MACHINES CORPORATION

Inventor(s)

Zvi Kons of Yoqneam Ilit (IL)

Aharon Satt of Kiryat Tivon (IL)

INTENT CLASSIFICATION ENHANCEMENT THROUGH TRAINING DATA AUGMENTATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17643224 titled 'INTENT CLASSIFICATION ENHANCEMENT THROUGH TRAINING DATA AUGMENTATION

Simplified Explanation

The abstract describes a computer-based method, system, and program that improves an intent classifier by augmenting training data. Here are the key points:

  • The method involves selecting a target sample from a group of samples.
  • An ambiguity level is determined for the target sample based on confidence scores of at least two intent labels associated with it.
  • If the ambiguity level is below a certain threshold, a nearest neighboring sample is selected from a group of neighboring samples.
  • The nearest neighboring sample has a confidence score associated with an intent label.
  • For each intent label, the confidence scores of the target sample and the nearest neighboring sample are merged into an overall confidence score.
  • The ambiguity level is modified using the overall confidence score.
  • The target sample is labeled with the intent label if the modified ambiguity level is above the threshold.

Potential applications of this technology:

  • Improving the accuracy and performance of intent classifiers in natural language processing systems.
  • Enhancing chatbots and virtual assistants to better understand user intents and provide more accurate responses.
  • Streamlining customer support systems by accurately categorizing and routing user queries.

Problems solved by this technology:

  • Addressing the challenge of accurately classifying user intents in natural language processing systems.
  • Overcoming the limitations of existing intent classifiers that may struggle with ambiguous or uncertain samples.
  • Improving the overall user experience by reducing misinterpretation of user intents.

Benefits of this technology:

  • Increased accuracy and reliability of intent classification in various applications.
  • Enhanced user satisfaction through improved understanding and response generation.
  • Time and cost savings by automating the classification process and reducing the need for manual intervention.


Original Abstract Submitted

A computer-implemented method, a computer system and a computer program product enhance an intent classifier through training data augmentation. The method includes selecting a target sample from a plurality of samples. The method also includes determining an ambiguity level for the target sample based on confidence scores of at least two intent labels associated with the target sample. The method further includes selecting a nearest neighboring sample from a group of neighboring samples when the ambiguity level is below a threshold. The nearest neighboring sample includes a confidence score associated with an intent label. The method also includes, for every intent label, merging the confidence scores of the two samples into an overall confidence score for the intent label and modifying the ambiguity level using the overall confidence score. Lastly, the method includes labeling the target sample with the intent label when the modified ambiguity level is above the threshold.

Retrieved from "http://wikipatents.org/index.php?title=17643224._INTENT_CLASSIFICATION_ENHANCEMENT_THROUGH_TRAINING_DATA_AUGMENTATION_simplified_abstract_(INTERNATIONAL_BUSINESS_MACHINES_CORPORATION)&oldid=25736"