Module document_summarizer 

Full-featured extractive and abstractive-style document summarization.

DocumentSummarizer implements five summarization strategies driven by TF-IDF, position bias, sentence length heuristics, and optional embedding centrality:

• Extractive — score every sentence and return the top-k in original order.
• Keyphrase — extract the most significant 2–4-word n-gram keyphrases.
• Headline — return the single most important sentence, truncated.
• Abstractive — concatenate top-3 sentences with transition words stripped, trimmed to a target word count.
• Hierarchical — cluster sentences and pick one representative per cluster.

Re-exports

pub use ds_types::DocumentChunk;

pub use ds_types::SentenceScore;

pub use ds_types::SummarizerConfig;

pub use ds_types::SummarizerError;

pub use ds_types::SummarizerStats;

pub use ds_types::SummaryResult;

pub use ds_types::SummaryStyle;

Modules

ds_types

Type definitions for the document summarizer.

Structs

DocumentSummarizer

Production-quality document summarizer supporting five summarization strategies.

Functions

cosine_similarity

Cosine similarity between two f64 slices; returns 0.0 on dimension mismatch or zero norm.

split_sentences

Split text into sentences on '. ', '! ', '? ', and '\n\n' boundaries.

tf_idf

Compute TF-IDF for term given the tokens of its document and the full corpus.

tokenize

Tokenize text into lowercase alphanumeric tokens.

xorshift64

Minimal xorshift64 PRNG; used in tests to avoid the rand crate.