1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
use std::ops::{Index, Range};
use std::str;
#[cfg(feature = "serde")]
use serde::{Deserialize, Serialize};
use smartstring::alias::String;
#[cfg(feature = "test-cases")]
pub mod test_cases;
#[cfg(feature = "__test_data")]
pub mod test_data;
#[cfg_attr(feature = "with-serde", derive(Deserialize, Serialize))]
pub struct Segmenter {
unigrams: HashMap<String, f64>,
bigrams: HashMap<(String, String), f64>,
uni_total: f64,
bi_total: f64,
limit: usize,
}
impl Segmenter {
pub fn from_iters<U, B>(unigrams: U, bigrams: B) -> Self
where
U: Iterator<Item = (String, f64)>,
B: Iterator<Item = ((String, String), f64)>,
{
Self::from_maps(unigrams.collect(), bigrams.collect())
}
pub fn from_maps(
unigrams: HashMap<String, f64>,
bigrams: HashMap<(String, String), f64>,
) -> Self {
Self {
uni_total: unigrams.values().sum(),
bi_total: bigrams.values().sum(),
unigrams,
bigrams,
limit: DEFAULT_LIMIT,
}
}
pub fn segment<'a>(
&self,
input: &str,
search: &'a mut Search,
) -> Result<impl Iterator<Item = &'a str> + ExactSizeIterator, InvalidCharacter> {
SegmentState::new(Ascii::new(input)?, &self, search).run();
Ok(search.result.iter().map(|v| v.as_str()))
}
pub fn score_sentence<'a>(&self, mut words: impl Iterator<Item = &'a str>) -> Option<f64> {
let mut prev = words.next()?;
let mut score = self.score(prev, None);
for word in words {
score += self.score(word, Some(prev));
prev = word;
}
Some(score)
}
fn score(&self, word: &str, previous: Option<&str>) -> f64 {
if let Some(prev) = previous {
if let Some(bi) = self.bigrams.get(&(prev.into(), word.into())) {
if let Some(uni) = self.unigrams.get(prev) {
return ((bi / self.bi_total) / (uni / self.uni_total)).log10();
}
}
}
match self.unigrams.get(word) {
Some(p) => p / self.uni_total,
None => 10.0 / (self.uni_total * 10.0f64.powi(word.len() as i32)),
}
.log10()
}
pub fn set_limit(&mut self, limit: usize) {
self.limit = limit;
}
}
struct SegmentState<'a> {
data: &'a Segmenter,
text: Ascii<'a>,
search: &'a mut Search,
}
impl<'a> SegmentState<'a> {
fn new(text: Ascii<'a>, data: &'a Segmenter, search: &'a mut Search) -> Self {
search.clear();
Self { data, text, search }
}
fn run(self) {
for end in 1..=self.text.len() {
let start = end.saturating_sub(self.data.limit);
for split in start..end {
let (prev, prev_score) = match split {
0 => (None, 0.0),
_ => {
let prefix = self.search.candidates[split - 1];
let word = &self.text[split - prefix.len as usize..split];
(Some(word), prefix.score)
}
};
let word = &self.text[split..end];
let score = self.data.score(word, prev) + prev_score;
match self.search.candidates.get_mut(end - 1) {
Some(cur) if cur.score < score => {
cur.len = end - split;
cur.score = score;
}
None => self.search.candidates.push(Candidate {
len: end - split,
score,
}),
_ => {}
}
}
}
let mut end = self.text.len();
let mut best = self.search.candidates[end - 1];
loop {
let word = &self.text[end - best.len as usize..end];
self.search.result.push(word.into());
end -= best.len as usize;
if end == 0 {
break;
}
best = self.search.candidates[end - 1];
}
self.search.result.reverse();
}
}
#[derive(Clone)]
pub struct Search {
candidates: Vec<Candidate>,
result: Vec<String>,
}
impl Search {
fn clear(&mut self) {
self.candidates.clear();
self.result.clear();
}
#[doc(hidden)]
pub fn get(&self, idx: usize) -> Option<&str> {
self.result.get(idx).map(|v| v.as_str())
}
}
impl Default for Search {
fn default() -> Self {
Self {
candidates: Vec::new(),
result: Vec::new(),
}
}
}
#[derive(Clone, Copy, Debug, Default)]
struct Candidate {
len: usize,
score: f64,
}
#[derive(Debug)]
struct Ascii<'a>(&'a [u8]);
impl<'a> Ascii<'a> {
fn new(s: &'a str) -> Result<Self, InvalidCharacter> {
let bytes = s.as_bytes();
match bytes.iter().all(|b| b.is_ascii_lowercase()) {
true => Ok(Self(bytes)),
false => Err(InvalidCharacter),
}
}
fn len(&self) -> usize {
self.0.len()
}
}
impl<'a> Index<Range<usize>> for Ascii<'a> {
type Output = str;
fn index(&self, index: Range<usize>) -> &Self::Output {
let bytes = self.0.index(index);
unsafe { str::from_utf8_unchecked(bytes) }
}
}
#[derive(Debug)]
pub struct InvalidCharacter;
impl std::error::Error for InvalidCharacter {}
impl std::fmt::Display for InvalidCharacter {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.write_str("invalid character")
}
}
type HashMap<K, V> = std::collections::HashMap<K, V, ahash::RandomState>;
const DEFAULT_LIMIT: usize = 24;
#[cfg(test)]
pub mod tests {
use super::*;
#[test]
fn test_clean() {
Ascii::new("Can't buy me love!").unwrap_err();
let text = Ascii::new("cantbuymelove").unwrap();
assert_eq!(&text[0..text.len()], "cantbuymelove");
}
}