Extractive summarization creates a summary by selecting relevant passages or sentences from a source and reproducing them verbatim, rather than generating new wording. section 1.3.1 section 9.0
The approach suits large-scale information workflows: transformer systems can summarize long documents and automatically curate news. Its selection process can also carry the priorities and biases present in the source history or training data. section 1.3.1 section 9.0
Verbatim code suggestions
An older code-generation system might find frequently occurring text in its history and reuse it as autocomplete—for example, suggesting “artificial intelligence” after “prosocial.” section 1.3.1
