WindowDiff Metric

A WindowDiff Metric is a text segmentation measure (for text segmentation).

• Context:
  • It can (typically) work with a sequence over a vocabulary of two items (e.g., "0", "1"), where a specified boundary value marks the edge of a segmentation.
  • It can (often) use a window width (k) as a parameter to determine the size of the comparison window.
  • It can measure the difference between two text segmentations by moving a fixed-size window across the text.
  • ...
• Example(s):
  • NLTK windowdiff(seg1, seg2, k, boundary='1').
  • ...
• Counter-Example(s):
  • ROUGE Score.
  • F1 Measure.
• See: Topic Segmentation, Natural Language Processing, NLTK.

---

---

References

2002

• (Pevzner & Hearst, 2002) ⇒ Lev Pevzner, Marti A. Hearst. (2002). “A Critique and Improvement of an Evaluation Metric for Text Segmentation.” In: Computational Linguistics, 28 (1).
  • QUOTE: “WindowDiff...is a measure used to compare two segmentations of a text by moving a fixed-sized window across the text and counting mismatches in the number of boundaries within the window.”

---