Is there any code to mark ”thought group" or "intonation unit" for English sentences?

[bk111]

I tried, but not good.
`import spacy

nlp = spacy.load("en_core_web_sm")

def get_ultra_fine_v2(text):
doc = nlp(text)
num_tokens = len(doc)
add_slash_after = [False] * num_tokens

# --- 1. 精确保护区 (仅保护绝对不可拆分的单位) ---
protected = [False] * num_tokens
# A. 保护冠词+名词 (a book), 限定词+名词 (my father)
for chunk in doc.noun_chunks:
    for i in range(chunk.start, chunk.end - 1):
        protected[i] = True
        
# B. 保护助动词+动词 (is standing, was playing)
for token in doc:
    if token.pos_ == "AUX" and token.i + 1 < num_tokens:
        if doc[token.i + 1].pos_ in ["VERB", "ADJ", "AUX"]:
            protected[token.i] = True
    if token.text.lower() == "to" and token.i + 1 < num_tokens and doc[token.i+1].pos_ == "VERB":
        protected[token.i] = True

# --- 2. 极致划分规则 (只要满足语义断层就尝试断句) ---
for i in range(num_tokens - 1):
    token = doc[i]
    next_token = doc[i+1]

    # 如果当前词是标点，直接跳过（标点自带断句效果）
    if token.pos_ == "PUNCT": continue

    # A. 引导词 (who, that, if, while...) 之前
    if next_token.text.lower() in ["who", "which", "that", "because", "while", "if", "when", "although"]:
        add_slash_after[i] = True

    # B. 介词短语 (in, for, at, with, over, of...) 之前
    # 只要前一个词不是动词（避免拆开 fix on 这种词组），就断开
    if next_token.pos_ == "ADP" and next_token.text.lower() not in ["to", "of"]:
        if token.pos_ not in ["VERB", "AUX", "DET", "ADJ"]:
            add_slash_after[i] = True

    # C. 副词短语/时间/地点状语 (over there, every day, hard...)
    if next_token.text.lower() in ["over", "every", "hard", "very"]:
        # 只有当它前面不是动词时（比如 standing / over there）
        add_slash_after[i] = True

    # D. 谓语动词/系动词 (is, was, are...) 之前
    # 只要主语部分多于 1 个词，就在谓语前断开
    if next_token.dep_ == "ROOT" or (next_token.pos_ == "AUX" and next_token.head.dep_ == "ROOT"):
        if token.pos_ not in ["PRON", "DET"]:
            add_slash_after[i] = True

    # E. 从句主语和谓语之间 (my sister / was playing)
    if next_token.pos_ in ["AUX", "VERB"] and token.pos_ in ["NOUN", "PROPN"]:
        add_slash_after[i] = True

# --- 3. 全局清洗 (排除不合理断句) ---
for i in range(num_tokens):
    # 保护区内的强制不断
    if protected[i]: add_slash_after[i] = False
    # 代词 (I, You, He)、冠词 (The, A) 之后不断
    if doc[i].pos_ in ["PRON", "DET"] and doc[i].text.lower() != "who": 
        add_slash_after[i] = False
    # 标点符号前、后不断
    if i + 1 < num_tokens and doc[i+1].pos_ == "PUNCT": add_slash_after[i] = False
    if doc[i].pos_ == "PUNCT": add_slash_after[i] = False

# --- 4. 组装 ---
result = []
for i in range(num_tokens):
    result.append(doc[i].text)
    if add_slash_after[i]:
        result.append("/")

output = " ".join(result)
for p in [",", ".", "!", "?", ":", ";"]:
    output = output.replace(f" {p}", p)
return output.strip()

--- 测试 ---

test_cases = [
"The man who is standing over there is my father.",
"If you want to succeed in your life, you must work hard every day.",
"English is a language that is spoken all over the world.",
"I believe that honesty is the best policy for everyone.",
"I was reading a book while my sister was playing the piano."
]

for s in test_cases:
print(f"\n[极致 2.0]: {get_ultra_fine_v2(s)}")`