SenseAudio 语音备忘录转录器

将语音备忘录转换为组织有序、可搜索的文本，并自动分类和提取关键信息。

此技能的功能

• - 将语音备忘录高精度转录为文本
• 将口语化表达转换为结构化书面格式
• 提取关键信息（日期、任务、联系人）
• 按主题或类别组织备忘录
• 生成摘要和待办事项

前置条件

安装所需的 Python 包：

bash
pip install requests

实施指南

第一步：转录语音备忘录

python
import requests

def transcribevoicememo(audio_file):
url = https://api.senseaudio.cn/v1/audio/transcriptions

headers = {Authorization: fBearer {API_KEY}}
files = {file: open(audio_file, rb)}
data = {
model: sense-asr, # 标准模型：功能完整，适合语音备忘录
response_format: json
}

response = requests.post(url, headers=headers, files=files, data=data)
return response.json()[text]

第二步：清理和结构化文本

将口语化表达转换为可读文本：

python
import re

def clean_transcription(text):
# 移除填充词
fillers = [um, uh, like, you know, basically, actually]
for filler in fillers:
text = re.sub(rf\b{filler}\b, , text, flags=re.IGNORECASE)

# 修正间距
text = re.sub(r\s+, , text).strip()

# 句子首字母大写
sentences = text.split(. )
text = . .join(s.capitalize() for s in sentences)

return text

第三步：提取关键信息

python
import re
from datetime import datetime

def extract_info(text):
info = {
dates: [],
tasks: [],
contacts: [],
keywords: []
}

# 提取日期
date_patterns = [
r\b(?:tomorrow|today|yesterday)\b,
r\b(?:monday|tuesday|wednesday|thursday|friday|saturday|sunday)\b,
r\b\d{1,2}/\d{1,2}/\d{2,4}\b
]
for pattern in date_patterns:
info[dates].extend(re.findall(pattern, text, re.IGNORECASE))

# 提取任务（动作动词）
task_patterns = [
r(?:need to|have to|must|should)\s+(\w+(?:\s+\w+){0,5}),
r(?:remember to|don\t forget to)\s+(\w+(?:\s+\w+){0,5})
]
for pattern in task_patterns:
info[tasks].extend(re.findall(pattern, text, re.IGNORECASE))

# 提取名称（大写单词）
info[contacts] = re.findall(r\b[A-Z][a-z]+(?:\s+[A-Z][a-z]+)*\b, text)

return info

第四步：分类备忘录

python
def categorize_memo(text):
categories = {
work: [meeting, project, deadline, client, email],
personal: [family, friend, home, weekend],
shopping: [buy, purchase, store, grocery],
ideas: [idea, think, maybe, could],
tasks: [todo, task, need to, must]
}

text_lower = text.lower()
scores = {}

for category, keywords in categories.items():
score = sum(1 for keyword in keywords if keyword in text_lower)
scores[category] = score

return max(scores, key=scores.get) if max(scores.values()) > 0 else general

第五步：生成结构化输出

python
def processvoicememo(audio_file):
# 转录
rawtext = transcribevoicememo(audiofile)

# 清理
cleantext = cleantranscription(raw_text)

# 提取信息
info = extractinfo(cleantext)

# 分类
category = categorizememo(cleantext)

# 创建结构化备忘录
memo = {
timestamp: datetime.now().isoformat(),
category: category,
text: clean_text,
rawtext: rawtext,
extracted_info: info,
summary: generatesummary(cleantext)
}

return memo

def generate_summary(text):
# 使用第一个句子或前100个字符
sentences = text.split(. )
return sentences[0] if sentences else text[:100]

高级功能

批量处理

处理多个备忘录：

python
def processmemobatch(audio_files):
memos = []
for audiofile in audiofiles:
memo = processvoicememo(audio_file)
memos.append(memo)

# 按类别分组
by_category = {}
for memo in memos:
category = memo[category]
if category not in by_category:
by_category[category] = []
by_category[category].append(memo)

return by_category

搜索和筛选

python
def search_memos(memos, query):
results = []
query_lower = query.lower()

for memo in memos:
if query_lower in memo[text].lower():
results.append(memo)

return results

def filterbydate(memos, date):
return [m for m in memos if date in m[extracted_info][dates]]

导出格式

python
def exporttomarkdown(memos):
md = # 语音备忘录\n\n

for memo in memos:
md += f## {memo[timestamp]}\n
md += f类别: {memo[category]}\n\n
md += f{memo[text]}\n\n

if memo[extracted_info][tasks]:
md += 任务:\n
for task in memo[extracted_info][tasks]:
md += f- [ ] {task}\n
md += \n

return md

输出格式

• - 清理后的转录文本
• 结构化的备忘录 JSON
• 提取的信息（日期、任务、联系人）
• 类别分类
• 摘要

最佳实践提示

• - 清晰并以正常语速说话
• 明确提及日期和姓名
• 使用动作动词表示任务
• 为获得最佳效果，备忘录控制在5分钟以内
• 检查并编辑提取的信息

参考

• - SenseAudio ASR API
• SenseAudio 平台