Text Summarization is a feature of machine learning software which can prepare a more concise version of a long text segment, so an attorney can scan through it and digest the key points more quickly. QuillBot offers a free online text summarization tool . See: https://quillbot.com/summarize .
There are two forms of text summarization. Extraction-based summarization pulls key terms and phrases from the analyzed text, without actually altering any of the original text. See this example:
Extraction-based summarization can be based on the length of keyphrases, and the frequency with which keyphrases occur. The keyphrases can be based on linguistic patterns of the text or be tagged by the reviewer.
Abstraction-based summarization re-phrases the source text, and prepares a more concise version, that aims to reflect the main points of the complete source text.
ROUGE, or Recall-Oriented Understudy for Gisting Evaluation, is software designed to evaluate text summaries by comparing them with example summaries that reviewers have determined to be accurate. ROGUE uses these techniques:
- Tracks the number of N-grams (words, syllables, or other sequences of text) that the source text has in common with the example that is confirmed to be accurate.
- Checks the longest common subsequences (LCS) between the source and the example.
- Factors in the number of consecutive longest common subsequences.
- Accounts for the co-occurrence of skip-bigrams: word pairs that are in the same sentence.