BatchalignHK 0.7.19.post10__tar.gz → 0.7.19.post12__tar.gz

{batchalignhk-0.7.19.post10 → batchalignhk-0.7.19.post12}/BatchalignHK.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: BatchalignHK
-Version: 0.7.19.post10
+Version: 0.7.19.post12
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

{batchalignhk-0.7.19.post10 → batchalignhk-0.7.19.post12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: BatchalignHK
-Version: 0.7.19.post10
+Version: 0.7.19.post12
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

{batchalignhk-0.7.19.post10 → batchalignhk-0.7.19.post12}/batchalign/cli/cli.py

@@ -331,6 +331,9 @@ def utseg(ctx, in_dir, out_dir, lang, num_speakers, **kwargs):
               show_default=True,
               default="eng",
               type=str)
+@click.option("--data",
+              help="the URL of the data",
+              type=str)
 @click.option("-n", "--num_speakers", type=int, help="number of speakers in the language sample", default=2)
 @click.pass_context
 def benchmark(ctx, in_dir, out_dir, lang, num_speakers, whisper, tencent, **kwargs):

{batchalignhk-0.7.19.post10 → batchalignhk-0.7.19.post12}/batchalign/pipelines/asr/tencent.py

@@ -111,6 +111,11 @@ class TencentEngine(BatchalignEngine):
             "重復": "重複"
         }
         return word_replacements.get(word, word)
+
+    @staticmethod
+    def is_roman(x):
+        """check if x contains only roman characters"""
+        return all(c.isalpha() and ord(c) < 128 for c in x if not c.isspace())
     
     def generate(self, f, **kwargs):
         lang = self.__lang
@@ -157,6 +162,9 @@ class TencentEngine(BatchalignEngine):
         for i in res.Data.ResultDetail:
             turn = []
             start = i.StartMs
+            roman_cache = ""
+            roman_cache_start = i.StartMs
+            roman_cache_end = i.StartMs
             for j in i.Words:
                 word = j.Word
                 if self.__lang == "yue":
@@ -164,12 +172,35 @@ class TencentEngine(BatchalignEngine):
 
                     word = self.replace_cantonese_words(word)
 
+                if self.is_roman(word):
+                    if roman_cache == "":
+                        roman_cache_start = (j.OffsetStartMs + start)
+                    roman_cache = roman_cache + word
+                    roman_cache_end = (j.OffsetEndMs + start)
+                else:
+                    if roman_cache != "":
+                        turn.append({
+                            "type": "text",
+                            "ts": roman_cache_start / 1000,
+                            "end_ts": roman_cache_end / 1000,
+                            "value": roman_cache
+                        })
+                    roman_cache = ""
+                    turn.append({
+                        "type": "text",
+                        "ts": (j.OffsetStartMs + start) / 1000,
+                        "end_ts": (j.OffsetEndMs + start) / 1000,
+                        "value": word
+                    })
+
+            if roman_cache != "":
                 turn.append({
                     "type": "text",
-                    "ts": (j.OffsetStartMs + start) / 1000,
-                    "end_ts": (j.OffsetEndMs + start) / 1000,
-                    "value": word
+                    "ts": roman_cache_start / 1000,
+                    "end_ts": roman_cache_end / 1000,
+                    "value": roman_cache
                 })
+
             turns.append({
                 "elements": turn,
                 "speaker": i.SpeakerId

batchalignhk-0.7.19.post12/batchalign/version

@@ -0,0 +1,3 @@
+0.7.19-post.12
+May 27th, 2025
+tencent: tokenize roman words together

batchalignhk-0.7.19.post10/batchalign/version

@@ -1,3 +0,0 @@
-0.7.19-post.10
-May 24th, 2025
-more asr changes