batchalign 0.7.12__tar.gz → 0.7.13.post1__tar.gz

{batchalign-0.7.12/batchalign.egg-info → batchalign-0.7.13.post1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: batchalign
-Version: 0.7.12
+Version: 0.7.13.post1
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

batchalign-0.7.13.post1/batchalign/pipelines/asr/num2chinese.py

@@ -0,0 +1,88 @@
+import itertools 
+
+def num2chinese(num, big=False, simp=True, o=False, twoalt=False):
+    """
+    Converts numbers to Chinese representations.
+
+    `big`   : use financial characters.
+    `simp`  : use simplified characters instead of traditional characters.
+    `o`     : use 〇 for zero.
+    `twoalt`: use 两/兩 for two when appropriate.
+
+    Note that `o` and `twoalt` is ignored when `big` is used, 
+    and `twoalt` is ignored when `o` is used for formal representations.
+    """
+
+    # check num first
+    nd = str(num)
+    if abs(float(nd)) >= 1e48:
+        raise ValueError('number out of range')
+    elif 'e' in nd:
+        raise ValueError('scientific notation is not supported')
+    c_symbol = '正负点' if simp else '正負點'
+    if o:  # formal
+        twoalt = False
+    if big:
+        c_basic = '零壹贰叁肆伍陆柒捌玖' if simp else '零壹貳參肆伍陸柒捌玖'
+        c_unit1 = '拾佰仟'
+        c_twoalt = '贰' if simp else '貳'
+    else:
+        c_basic = '〇一二三四五六七八九' if o else '零一二三四五六七八九'
+        c_unit1 = '十百千'
+        if twoalt:
+            c_twoalt = '两' if simp else '兩'
+        else:
+            c_twoalt = '二'
+    c_unit2 = '万亿兆京垓秭穰沟涧正载' if simp else '萬億兆京垓秭穰溝澗正載'
+    revuniq = lambda l: ''.join(k for k, g in itertools.groupby(reversed(l)))
+    nd = str(num)
+    result = []
+    if nd[0] == '+':
+        result.append(c_symbol[0])
+    elif nd[0] == '-':
+        result.append(c_symbol[1])
+    if '.' in nd:
+        integer, remainder = nd.lstrip('+-').split('.')
+    else:
+        integer, remainder = nd.lstrip('+-'), None
+    if int(integer):
+        splitted = [integer[max(i - 4, 0):i]
+                    for i in range(len(integer), 0, -4)]
+        intresult = []
+        for nu, unit in enumerate(splitted):
+            # special cases
+            if int(unit) == 0:  # 0000
+                intresult.append(c_basic[0])
+                continue
+            elif nu > 0 and int(unit) == 2:  # 0002
+                intresult.append(c_twoalt + c_unit2[nu - 1])
+                continue
+            ulist = []
+            unit = unit.zfill(4)
+            for nc, ch in enumerate(reversed(unit)):
+                if ch == '0':
+                    if ulist:  # ???0
+                        ulist.append(c_basic[0])
+                elif nc == 0:
+                    ulist.append(c_basic[int(ch)])
+                elif nc == 1 and ch == '1' and unit[1] == '0':
+                    # special case for tens
+                    # edit the 'elif' if you don't like
+                    # 十四, 三千零十四, 三千三百一十四
+                    ulist.append(c_unit1[0])
+                elif nc > 1 and ch == '2':
+                    ulist.append(c_twoalt + c_unit1[nc - 1])
+                else:
+                    ulist.append(c_basic[int(ch)] + c_unit1[nc - 1])
+            ustr = revuniq(ulist)
+            if nu == 0:
+                intresult.append(ustr)
+            else:
+                intresult.append(ustr + c_unit2[nu - 1])
+        result.append(revuniq(intresult).strip(c_basic[0]))
+    else:
+        result.append(c_basic[0])
+    if remainder:
+        result.append(c_symbol[2])
+        result.append(''.join(c_basic[int(ch)] for ch in remainder))
+    return ''.join(result)

{batchalign-0.7.12 → batchalign-0.7.13.post1}/batchalign/pipelines/asr/utils.py

@@ -3,6 +3,7 @@ from batchalign.document import *
 from batchalign.utils import *
 
 from batchalign.constants import ENDING_PUNCT
+from batchalign.pipelines.asr.num2chinese import num2chinese
 
 from num2words import num2words
 import pycountry
@@ -165,7 +166,17 @@ def process_generation(output, lang="eng", utterance_engine=None):
             try:
                 return num2words(i, lang=lang_2)
             except NotImplementedError:
-                return i
+                try:
+                    if lang == "zho":
+                        return num2chinese(i)
+                    elif lang == "jpn":
+                        return num2chinese(i, simp=False)
+                    elif lang == "yue":
+                        return num2chinese(i, simp=False)
+                    else:
+                        return i
+                except:
+                    return i
         final_words = [[catched_num2words(i), j] for i,j in final_words]
 
         # if the final words is > 300, split into n parts

{batchalign-0.7.12 → batchalign-0.7.13.post1}/batchalign/pipelines/morphosyntax/ud.py

@@ -826,6 +826,7 @@ def morphoanalyze(doc: Document, retokenize:bool, status_hook:callable = None, *
         line_cut = line_cut.replace("+//", "")
         line_cut = line_cut.replace("+...", "")
         line_cut = line_cut.replace("_", "")
+        line_cut = line_cut.replace("#", "")
 
         # xbxxx is a sepecial xxx-class token to mark
         # special form markers, used for processing later

batchalign-0.7.13.post1/batchalign/version

@@ -0,0 +1,3 @@
+0.7.13-post.1
+Feburary 14nd, 2025
+Remove hash sign.

{batchalign-0.7.12 → batchalign-0.7.13.post1/batchalign.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: batchalign
-Version: 0.7.12
+Version: 0.7.13.post1
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

{batchalign-0.7.12 → batchalign-0.7.13.post1}/batchalign.egg-info/SOURCES.txt

@@ -57,6 +57,7 @@ batchalign/pipelines/pipeline.py
 batchalign/pipelines/analysis/__init__.py
 batchalign/pipelines/analysis/eval.py
 batchalign/pipelines/asr/__init__.py
+batchalign/pipelines/asr/num2chinese.py
 batchalign/pipelines/asr/rev.py
 batchalign/pipelines/asr/utils.py
 batchalign/pipelines/asr/whisper.py

batchalign-0.7.12/batchalign/version

@@ -1,3 +0,0 @@
-0.7.12
-Feburary 6nd, 2025
-Wav2vec support!