gardel/Mofox-Core
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

better：优化interest的算法，更好更强

Browse Source
This commit is contained in:
SengokuCola
2025-07-25 15:45:14 +08:00
parent 2aec68bd3d
commit 8c9b2b54c0
4 changed files with 511 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
src/chat/heart_flow/heartflow_message_processor.py
View File

@@ -61,11 +61,35 @@ async def _calculate_interest(message: MessageRecv) -> Tuple[float, bool]:
logger.debug(f"记忆激活率: {interested_rate:.2f}")
text_len = len(message.processed_plain_text)
# 根据文本长度调整兴趣度，长度越大兴趣度越高，但增长率递减，最低0.01，最高0.05
# 采用对数函数实现递减增长
base_interest = 0.01 + (0.05 - 0.01) * (math.log10(text_len + 1) / math.log10(1000 + 1))
base_interest = min(max(base_interest, 0.01), 0.05)
# 根据文本长度分布调整兴趣度，采用分段函数实现更精确的兴趣度计算
# 基于实际分布：0-5字符(26.57%), 6-10字符(27.18%), 11-20字符(22.76%), 21-30字符(10.33%), 31+字符(13.86%)
if text_len == 0:
base_interest = 0.01 # 空消息最低兴趣度
elif text_len <= 5:
# 1-5字符：线性增长 0.01 -> 0.03
base_interest = 0.01 + (text_len - 1) * (0.03 - 0.01) / 4
elif text_len <= 10:
# 6-10字符：线性增长 0.03 -> 0.06
base_interest = 0.03 + (text_len - 5) * (0.06 - 0.03) / 5
elif text_len <= 20:
# 11-20字符：线性增长 0.06 -> 0.12
base_interest = 0.06 + (text_len - 10) * (0.12 - 0.06) / 10
elif text_len <= 30:
# 21-30字符：线性增长 0.12 -> 0.18
base_interest = 0.12 + (text_len - 20) * (0.18 - 0.12) / 10
elif text_len <= 50:
# 31-50字符：线性增长 0.18 -> 0.22
base_interest = 0.18 + (text_len - 30) * (0.22 - 0.18) / 20
elif text_len <= 100:
# 51-100字符：线性增长 0.22 -> 0.26
base_interest = 0.22 + (text_len - 50) * (0.26 - 0.22) / 50
else:
# 100+字符：对数增长 0.26 -> 0.3，增长率递减
base_interest = 0.26 + (0.3 - 0.26) * (math.log10(text_len - 99) / math.log10(901)) # 1000-99=901
# 确保在范围内
base_interest = min(max(base_interest, 0.01), 0.3)
interested_rate += base_interest

124
src/chat/memory_system/Hippocampus.py
View File

@@ -299,6 +299,63 @@ class Hippocampus:
# 按相似度降序排序
memories.sort(key=lambda x: x[2], reverse=True)
return memories
async def get_keywords_from_text(self, text: str, fast_retrieval: bool = False) -> list:
"""从文本中提取关键词。
Args:
text (str): 输入文本
fast_retrieval (bool, optional): 是否使用快速检索。默认为False。
如果为True，使用jieba分词提取关键词，速度更快但可能不够准确。
如果为False，使用LLM提取关键词，速度较慢但更准确。
"""
if not text:
return []
if fast_retrieval:
# 使用jieba分词提取关键词
words = jieba.cut(text)
# 过滤掉停用词和单字词
keywords = [word for word in words if len(word) > 1]
# 去重
keywords = list(set(keywords))
# 限制关键词数量
logger.debug(f"提取关键词: {keywords}")
else:
# 使用LLM提取关键词 - 根据详细文本长度分布优化topic_num计算
text_length = len(text)
if text_length <= 5:
topic_num = 1 # 1-5字符: 1个关键词 (26.57%的文本)
elif text_length <= 10:
topic_num = 1 # 6-10字符: 1个关键词 (27.18%的文本)
elif text_length <= 20:
topic_num = 2 # 11-20字符: 2个关键词 (22.76%的文本)
elif text_length <= 30:
topic_num = 3 # 21-30字符: 3个关键词 (10.33%的文本)
elif text_length <= 50:
topic_num = 4 # 31-50字符: 4个关键词 (9.79%的文本)
else:
topic_num = 5 # 51+字符: 5个关键词 (其余长文本)
# logger.info(f"提取关键词数量: {topic_num}")
topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
self.find_topic_llm(text, topic_num)
)
# 提取关键词
keywords = re.findall(r"<([^>]+)>", topics_response)
if not keywords:
keywords = []
else:
keywords = [
keyword.strip()
for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
if keyword.strip()
]
return keywords
async def get_memory_from_text(
self,
@@ -325,39 +382,7 @@ class Hippocampus:
- memory_items: list, 该主题下的记忆项列表
- similarity: float, 与文本的相似度
"""
if not text:
return []
if fast_retrieval:
# 使用jieba分词提取关键词
words = jieba.cut(text)
# 过滤掉停用词和单字词
keywords = [word for word in words if len(word) > 1]
# 去重
keywords = list(set(keywords))
# 限制关键词数量
logger.debug(f"提取关键词: {keywords}")
else:
# 使用LLM提取关键词
topic_num = min(5, max(1, int(len(text) * 0.1))) # 根据文本长度动态调整关键词数量
# logger.info(f"提取关键词数量: {topic_num}")
topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
self.find_topic_llm(text, topic_num)
)
# 提取关键词
keywords = re.findall(r"<([^>]+)>", topics_response)
if not keywords:
keywords = []
else:
keywords = [
keyword.strip()
for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
if keyword.strip()
]
# logger.info(f"提取的关键词: {', '.join(keywords)}")
keywords = await self.get_keywords_from_text(text, fast_retrieval)
# 过滤掉不存在于记忆图中的关键词
valid_keywords = [keyword for keyword in keywords if keyword in self.memory_graph.G]
@@ -679,38 +704,7 @@ class Hippocampus:
Returns:
float: 激活节点数与总节点数的比值
"""
if not text:
return 0
if fast_retrieval:
# 使用jieba分词提取关键词
words = jieba.cut(text)
# 过滤掉停用词和单字词
keywords = [word for word in words if len(word) > 1]
# 去重
keywords = list(set(keywords))
# 限制关键词数量
keywords = keywords[:5]
else:
# 使用LLM提取关键词
topic_num = min(5, max(1, int(len(text) * 0.1))) # 根据文本长度动态调整关键词数量
# logger.info(f"提取关键词数量: {topic_num}")
topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
self.find_topic_llm(text, topic_num)
)
# 提取关键词
keywords = re.findall(r"<([^>]+)>", topics_response)
if not keywords:
keywords = []
else:
keywords = [
keyword.strip()
for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
if keyword.strip()
]
# logger.info(f"提取的关键词: {', '.join(keywords)}")
keywords = await self.get_keywords_from_text(text, fast_retrieval)
# 过滤掉不存在于记忆图中的关键词
valid_keywords = [keyword for keyword in keywords if keyword in self.memory_graph.G]

34
src/mais4u/mais4u_chat/s4u_msg_processor.py
View File

@@ -47,11 +47,35 @@ async def _calculate_interest(message: MessageRecv) -> Tuple[float, bool]:
logger.debug(f"记忆激活率: {interested_rate:.2f}")
text_len = len(message.processed_plain_text)
# 根据文本长度调整兴趣度，长度越大兴趣度越高，但增长率递减，最低0.01，最高0.05
# 采用对数函数实现递减增长
base_interest = 0.01 + (0.05 - 0.01) * (math.log10(text_len + 1) / math.log10(1000 + 1))
base_interest = min(max(base_interest, 0.01), 0.05)
# 根据文本长度分布调整兴趣度，采用分段函数实现更精确的兴趣度计算
# 基于实际分布：0-5字符(26.57%), 6-10字符(27.18%), 11-20字符(22.76%), 21-30字符(10.33%), 31+字符(13.86%)
if text_len == 0:
base_interest = 0.01 # 空消息最低兴趣度
elif text_len <= 5:
# 1-5字符：线性增长 0.01 -> 0.03
base_interest = 0.01 + (text_len - 1) * (0.03 - 0.01) / 4
elif text_len <= 10:
# 6-10字符：线性增长 0.03 -> 0.06
base_interest = 0.03 + (text_len - 5) * (0.06 - 0.03) / 5
elif text_len <= 20:
# 11-20字符：线性增长 0.06 -> 0.12
base_interest = 0.06 + (text_len - 10) * (0.12 - 0.06) / 10
elif text_len <= 30:
# 21-30字符：线性增长 0.12 -> 0.18
base_interest = 0.12 + (text_len - 20) * (0.18 - 0.12) / 10
elif text_len <= 50:
# 31-50字符：线性增长 0.18 -> 0.22
base_interest = 0.18 + (text_len - 30) * (0.22 - 0.18) / 20
elif text_len <= 100:
# 51-100字符：线性增长 0.22 -> 0.26
base_interest = 0.22 + (text_len - 50) * (0.26 - 0.22) / 50
else:
# 100+字符：对数增长 0.26 -> 0.3，增长率递减
base_interest = 0.26 + (0.3 - 0.26) * (math.log10(text_len - 99) / math.log10(901)) # 1000-99=901
# 确保在范围内
base_interest = min(max(base_interest, 0.01), 0.3)
interested_rate += base_interest