unobtanium_segmenter/segmentation/
unicode_sentence.rs

1use unicode_segmentation::UnicodeSegmentation;
2
3use std::vec::IntoIter;
4
5use crate::SegmentedToken;
6use crate::UseOrSubdivide;
7
8use crate::segmentation::Segmenter;
9
10/// Split Large chunks of text into sentances according to the Unicode definition of a Sentence.
11///
12/// Uses [[UnicodeSegmentation::split_sentence_bounds]] under the hood.
13#[derive(Debug, Clone, Default)]
14pub struct UnicodeSentenceSplitter {}
15
16impl UnicodeSentenceSplitter {
17	/// Create a new UnicodeSentenceSplitter instance
18	pub fn new() -> Self {
19		Default::default()
20	}
21}
22
23impl Segmenter for UnicodeSentenceSplitter {
24	type SubdivisionIter<'a> = IntoIter<SegmentedToken<'a>>;
25
26	fn subdivide<'a>(
27		&self,
28		token: SegmentedToken<'a>,
29	) -> UseOrSubdivide<SegmentedToken<'a>, IntoIter<SegmentedToken<'a>>> {
30		let sentence_iterator = token.text.split_sentence_bounds();
31		let mut collection = Vec::new();
32		for sentence in sentence_iterator {
33			collection.push(SegmentedToken::new_derived_from(sentence, &token));
34		}
35		if collection.len() > 1 {
36			return UseOrSubdivide::Subdivide(collection.into_iter());
37		} else {
38			return UseOrSubdivide::Use(token);
39		}
40	}
41}
42
43#[cfg(test)]
44mod test {
45
46	use super::*;
47
48	use crate::chain::ChainSegmenter;
49	use crate::chain::StartSegmentationChain;
50
51	#[test]
52	fn test_unicode_sentence_split() {
53		let test_text = "The quick (\"brown\") fox can't jump 32.3 feet, right?\nThe quick (\"brown\")  fox. The value of π in german is '3,141592…'.";
54
55		let result: Vec<&str> = test_text
56			.start_segmentation_chain()
57			.chain_segmenter(&UnicodeSentenceSplitter::new())
58			.map(|t| t.text)
59			.collect();
60
61		let expected_tokens = vec![
62			"The quick (\"brown\") fox can't jump 32.3 feet, right?\n",
63			"The quick (\"brown\")  fox. ",
64			"The value of π in german is '3,141592…'.",
65		];
66
67		assert_eq!(result, expected_tokens);
68	}
69}
unobtanium_segmenter/segmentation/unicode_sentence.rs

unobtanium_segmenter/segmentation/
unicode_sentence.rs
