Upload 3 files

app.py

@@ -1,64 +1,19 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+from utils import analyze_chat
+
+# 创建Gradio界面
+iface = gr.Interface(
+    fn=analyze_chat,
+    inputs=gr.File(label="上传微信聊天记录（JSON / TXT）"),
+    outputs=[
+        gr.Textbox(label="实体识别结果", lines=4),
+        gr.Textbox(label="人物关系（含关系类型）", lines=6),
+        gr.HTML(label="人物关系图谱")
     ],
+    title="微信聊天人物关系分析",
+    description="上传JSON或TXT格式的微信聊天记录，自动识别人物实体和关系类型，并生成互动图谱。",
+    theme="compact"
 )
 
-
 if __name__ == "__main__":
-    demo.launch()
+    iface.launch()

requirements.txt

@@ -1 +1,5 @@
-huggingface_hub==0.25.2
+gradio==4.25.0
+transformers>=4.40.0
+torch>=2.1.0
+pyvis
+networkx

utils.py

@@ -0,0 +1,147 @@
+import json
+import tempfile
+import re
+from collections import defaultdict
+
+from transformers import (
+    AutoTokenizer,
+    AutoModelForTokenClassification,
+    AutoModelForSequenceClassification,
+    pipeline,
+)
+import torch
+from pyvis.network import Network
+
+
+# -------------------------------
+# 实体识别模型（NER）
+# -------------------------------
+ner_tokenizer = AutoTokenizer.from_pretrained("ckiplab/bert-base-chinese-ner")
+ner_model = AutoModelForTokenClassification.from_pretrained("ckiplab/bert-base-chinese-ner")
+ner_pipeline = pipeline("ner", model=ner_model, tokenizer=ner_tokenizer, aggregation_strategy="simple")
+
+
+# -------------------------------
+# 人物关系分类模型（BERT 分类器）
+# -------------------------------
+rel_model_name = "uer/roberta-base-finetuned-baike-chinese-relation-extraction"
+rel_tokenizer = AutoTokenizer.from_pretrained(rel_model_name)
+rel_model = AutoModelForSequenceClassification.from_pretrained(rel_model_name)
+rel_model.eval()
+
+id2label = {
+    0: "夫妻", 1: "父子", 2: "朋友", 3: "师生", 4: "同事", 5: "其他"
+}
+
+
+def classify_relation_bert(e1, e2, context):
+    prompt = f"{e1}和{e2}的关系是？{context}"
+    inputs = rel_tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512)
+    with torch.no_grad():
+        logits = rel_model(**inputs).logits
+    pred = torch.argmax(logits, dim=1).item()
+    probs = torch.nn.functional.softmax(logits, dim=1)
+    confidence = probs[0, pred].item()
+    return f"{id2label[pred]}（置信度 {confidence:.2f}）"
+
+
+# -------------------------------
+# 聊天输入解析
+# -------------------------------
+def parse_input_file(file):
+    filename = file.name
+    if filename.endswith(".json"):
+        return json.load(file)
+    elif filename.endswith(".txt"):
+        content = file.read().decode("utf-8")
+        lines = content.strip().splitlines()
+        chat_data = []
+        for line in lines:
+            match = re.match(r"(\d{4}-\d{2}-\d{2}.*?) (.*?): (.*)", line)
+            if match:
+                _, sender, message = match.groups()
+                chat_data.append({"sender": sender, "message": message})
+        return chat_data
+    else:
+        raise ValueError("不支持的文件格式，请上传 JSON 或 TXT 文件")
+
+
+# -------------------------------
+# 实体提取函数
+# -------------------------------
+def extract_entities(text):
+    results = ner_pipeline(text)
+    people = set()
+    for r in results:
+        if r["entity_group"] == "PER":
+            people.add(r["word"])
+    return list(people)
+
+
+# -------------------------------
+# 关系抽取函数（共现 + BERT 分类）
+# -------------------------------
+def extract_relations(chat_data, entities):
+    relations = defaultdict(lambda: defaultdict(lambda: {"count": 0, "contexts": []}))
+
+    for entry in chat_data:
+        msg = entry["message"]
+        found = [e for e in entities if e in msg]
+        for i in range(len(found)):
+            for j in range(i + 1, len(found)):
+                e1, e2 = found[i], found[j]
+                relations[e1][e2]["count"] += 1
+                relations[e1][e2]["contexts"].append(msg)
+                relations[e2][e1]["count"] += 1
+                relations[e2][e1]["contexts"].append(msg)
+
+    edges = []
+    for e1 in relations:
+        for e2 in relations[e1]:
+            if e1 < e2:
+                context_text = " ".join(relations[e1][e2]["contexts"])
+                label = classify_relation_bert(e1, e2, context_text)
+                edges.append((e1, e2, relations[e1][e2]["count"], label))
+    return edges
+
+
+# -------------------------------
+# 图谱绘制
+# -------------------------------
+def draw_graph(entities, relations):
+    g = Network(height="600px", width="100%", notebook=False)
+    g.barnes_hut()
+    for ent in entities:
+        g.add_node(ent, label=ent)
+    for e1, e2, weight, label in relations:
+        g.add_edge(e1, e2, value=weight, title=f"{label}（互动{weight}次）", label=label)
+    tmp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".html")
+    g.show(tmp_file.name)
+    with open(tmp_file.name, 'r', encoding='utf-8') as f:
+        return f.read()
+
+
+# -------------------------------
+# 主流程函数
+# -------------------------------
+def analyze_chat(file):
+    if file is None:
+        return "请上传聊天文件", "", ""
+    
+    try:
+        content = parse_input_file(file)
+    except Exception as e:
+        return f"读取文件失败: {e}", "", ""
+
+    text = "\n".join([entry["sender"] + ": " + entry["message"] for entry in content])
+    entities = extract_entities(text)
+    if not entities:
+        return "未识别到任何人物实体", "", ""
+
+    relations = extract_relations(content, entities)
+    if not relations:
+        return "未发现人物之间的关系", "", ""
+
+    graph_html = draw_graph(entities, relations)
+
+    return str(entities), str(relations), graph_html