Merge pull request #45 from tuna2134/number

数字を読めれるように

Cargo.lock

@@ -1361,9 +1361,9 @@ dependencies = [
 
 [[package]]
 name = "once_cell"
-version = "1.19.0"
+version = "1.20.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3fdb12b2476b595f9358c5161aa467c2438859caa136dec86c26fdd2efe17b92"
+checksum = "33ea5043e58958ee56f3e15a90aee535795cd7dfd319846288d93c5b57d85cbe"
 
 [[package]]
 name = "onig"

Cargo.toml

@@ -6,4 +6,4 @@ members = ["sbv2_api", "sbv2_core", "sbv2_bindings"]
 anyhow = "1.0.86"
 dotenvy = "0.15.7"
 env_logger = "0.11.5"
-ndarray = "0.16.1"
+ndarray = "0.16.1"

content.txt

@@ -1,7 +1 @@
-日本語を母国語としない人々にとって、「日本語は非常に難しい言語である」と言われています。
-その理由として、
-・漢字、ひらがな、カタカナ、と表記が何種類もある。
-・同一の漢字でも音読みと訓読みがある
-・地名の読みが難しい
-・主語、述語が省略される
-などが挙げられます。
+10,000年前までコロナが流行っていました

sbv2_core/src/jtalk.rs

@@ -54,6 +54,17 @@ impl JTalk {
         Ok(Self { jpreprocess })
     }
 
+    pub fn num2word(&self, text: &str) -> Result<String> {
+        let mut parsed = self.jpreprocess.text_to_njd(text)?;
+        parsed.preprocess();
+        let texts: Vec<String> = parsed
+            .nodes
+            .iter()
+            .map(|x| x.get_string().to_string())
+            .collect();
+        Ok(texts.join(""))
+    }
+
     pub fn process_text(&self, text: &str) -> Result<JTalkProcess> {
         let parsed = self.jpreprocess.run_frontend(text)?;
         let jtalk_process = JTalkProcess::new(Arc::clone(&self.jpreprocess), parsed);

sbv2_core/src/norm.rs

@@ -120,7 +120,8 @@ pub fn replace_punctuation(mut text: String) -> String {
     for (k, v) in REPLACE_MAP.iter() {
         text = text.replace(k, v);
     }
-    PUNCTUATION_CLEANUP_PATTERN
+    let content = PUNCTUATION_CLEANUP_PATTERN
         .replace_all(&text, "")
-        .to_string()
+        .to_string();
+    content
 }

sbv2_core/src/tts.rs

@@ -123,7 +123,8 @@ impl TTSModelHolder {
         &self,
         text: &str,
     ) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
-        let normalized_text = norm::normalize_text(text);
+        let text = self.jtalk.num2word(text)?;
+        let normalized_text = norm::normalize_text(&text);
 
         let process = self.jtalk.process_text(&normalized_text)?;
         let (phones, tones, mut word2ph) = process.g2p()?;