News Segmentation Results

The following XML shows a single record produced by news segmentation.

<NewsSegmentationResult>
  <id>af4475f5-beb2-401e-bf73-fb494f63af27</id>
  <storyText>business secretary says the government will consider co investing
             with a Buy on commercial terms <SIL> ...</storyText>
  <score>0</score>
  <terms>
    <term>
      <text>Buy</text>
      <weight>170</weight>
    </term>
    <term>
      <text>business secretary says</text>
      <weight>120</weight>
    </term>
    <term>
      <text>consider co investing</text>
      <weight>110</weight>
    </term>
    <term>
      <text>commercial terms</text>
      <weight>90</weight>
    </term>
    ...
  </terms>
  <type>Short story</type>
</NewsSegmentationResult>

The record contains the following elements:

  • storyText contains the text extracted from the video. The text is extracted from the audio by speech-to-text. Words are then combined into segments by the text segmentation task. News Segmentation analyzes the segments and combines one or more segments into a result that represents a story, short story, or video with no topic.
  • score is an integer from 0 to 100 that indicates the consistency between the terms in each segment that makes up the storyText.
  • each terms/term/text element contains a key term extracted from the text.
  • each terms/term/weight element contains the weight of the associated term. You might want to use this weight when building a term cloud. The weights are relative to the other terms in the story, and cannot be compared across stories.
  • the type element specifies whether the news segmentation task has classified the segment as a Story, Short Story, or a segment with No topic.