File size: 4,270 Bytes
6378099
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4f56acc
6378099
 
 
 
 
 
169746e
6378099
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
169746e
32563c3
169746e
6378099
 
9e6cedb
32563c3
 
 
33860c8
6378099
32563c3
 
 
6378099
 
 
 
 
 
 
 
 
 
 
 
 
33860c8
32563c3
 
 
 
 
 
 
9e6cedb
32563c3
 
9e6cedb
6378099
 
 
 
32563c3
6378099
9e6cedb
32563c3
 
9e6cedb
32563c3
 
 
 
 
 
9e6cedb
32563c3
9e6cedb
32563c3
 
9e6cedb
32563c3
 
 
9e6cedb
32563c3
 
9e6cedb
4f56acc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
# from flask import Flask, request, jsonify
# from sentence_transformers import CrossEncoder



# app = Flask(__name__)


# # Load your cross-encoder model
# model_name = "truong1301/reranker_pho_BLAI"  # Replace with your actual model if different
# cross_encoder = CrossEncoder(model_name, max_length=256, num_labels=1)

# # Function to preprocess text with Vietnamese word segmentation
# def preprocess_text(text):
#     if not text:
#         return text
#     segmented_text = rdrsegmenter.word_segment(text)
#     # Join tokenized sentences into a single string
#     return " ".join([" ".join(sentence) for sentence in segmented_text])

# @app.route("/rerank", methods=["POST"])
# def rerank():
#     try:
#         # Get JSON data from the request (query and list of documents)
#         data = request.get_json()
#         query = data.get("query", "")
#         documents = data.get("documents", [])

#         if not query or not documents:
#             return jsonify({"error": "Missing query or documents"}), 400

#         # Create pairs of query and documents for reranking
#         query_doc_pairs = [(query, doc) for doc in documents]

#         # Get reranking scores from the cross-encoder
#         scores = cross_encoder.predict(query_doc_pairs).tolist()

#         # Combine documents with their scores and sort
#         ranked_results = sorted(
#             [{"document": doc, "score": score} for doc, score in zip(documents, scores)],
#             key=lambda x: x["score"],
#             reverse=True
#         )

#         return jsonify({"results": ranked_results})

#     except Exception as e:
#         return jsonify({"error": str(e)}), 500

# @app.route("/", methods=["GET"])
# def health_check():
#     return jsonify({"status": "Server is running"}), 200

# if __name__ == "__main__":
#     app.run(host="0.0.0.0", port=7860)  # Default port for Hugging Face Spaces


from flask import Flask, request, jsonify
from transformers import pipeline
from sentence_transformers import CrossEncoder

app = Flask(__name__)

# Load Vietnamese word segmentation pipeline
segmenter = pipeline("token-classification", model="NlpHUST/vi-word-segmentation")

# Load your cross-encoder model
model_name = "truong1301/reranker_pho_BLAI"  # Replace with your actual model if different
cross_encoder = CrossEncoder(model_name, max_length=256, num_labels=1)

# Function to preprocess text using Vietnamese word segmentation
def preprocess_text(text):
    if not text:
        return text

    ner_results = segmenter(text)
    segmented_text = ""
    
    for e in ner_results:
        if "##" in e["word"]:
            segmented_text += e["word"].replace("##", "")
        elif e["entity"] == "I":
            segmented_text += "_" + e["word"]
        else:
            segmented_text += " " + e["word"]

    return segmented_text.strip()

@app.route("/rerank", methods=["POST"])
def rerank():
    try:
        # Get JSON data from the request (query and list of documents)
        data = request.get_json()
        query = data.get("query", "")
        documents = data.get("documents", [])

        if not query or not documents:
            return jsonify({"error": "Missing query or documents"}), 400

        # Apply Vietnamese word segmentation preprocessing
        segmented_query = preprocess_text(query)
        segmented_documents = [preprocess_text(doc) for doc in documents]

        # Create pairs of query and documents for reranking
        query_doc_pairs = [(segmented_query, doc) for doc in segmented_documents]

        # Get reranking scores from the cross-encoder
        scores = cross_encoder.predict(query_doc_pairs).tolist()

        # Combine documents with their scores and sort
        ranked_results = sorted(
            [{"document": doc, "score": score} for doc, score in zip(documents, scores)],
            key=lambda x: x["score"],
            reverse=True
        )

        return jsonify({"results": ranked_results})

    except Exception as e:
        return jsonify({"error": str(e)}), 500

@app.route("/", methods=["GET"])
def health_check():
    return jsonify({"status": "Server is running"}), 200

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=7860)  # Default port for Hugging Face Spaces