METHODS TO REDUCE SKIP RATES IN SPEECH DATA LABELING

Abstract

This invention proposes a method to reduce the skip rate in speech data labeling, implemented through the following steps: Step 1: Collect speech segments to be labeled with text; Step 2: Label the text; Step 3: Create a training set for a machine learning model; Step 4: Build the machine learning model; Step 5: Train the machine learning model; Step 6: Use the machine learning model to filter the data. The method helps reduce time and increase productivity in speech data labeling while ensuring data quality. The method uses a machine learning model to learn the skipping behavior of the labeler, thereby eliminating potentially skippable speech segments before providing the data to the labeler.