indiejoseph commited on
Commit
4260211
·
verified ·
1 Parent(s): f1249b7

Upload 3 files

Browse files
Files changed (4) hide show
  1. .gitattributes +1 -0
  2. app.py +97 -0
  3. emb.json +3 -0
  4. requirements.txt +2 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ emb.json filter=lfs diff=lfs merge=lfs -text
app.py ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import pandas as pd
2
+ import numpy as np
3
+ import asyncio
4
+ from sentence_transformers import SentenceTransformer
5
+
6
+ model = SentenceTransformer("hon9kon9ize/bert-large-cantonese")
7
+ df = None
8
+
9
+
10
+ def get_dataframe():
11
+ global df
12
+
13
+ if df is not None:
14
+ return df
15
+
16
+ df = pd.read_json("emb.json")
17
+ df = df.drop_duplicates(subset=["artist_name", "track_name"])
18
+
19
+ return df
20
+
21
+
22
+ def cosine_similarity(a, b):
23
+ return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))
24
+
25
+
26
+ def find_songs(text):
27
+ df = get_dataframe()
28
+ text_embedding = get_embedding(text)
29
+ df["similarity"] = df["lyrics_embedding"].apply(
30
+ lambda x: cosine_similarity(x, text_embedding)
31
+ )
32
+ df = df.sort_values(by="similarity", ascending=False)
33
+ top_5 = df.head(5)
34
+
35
+ return "### 以下係你嘅歌名推介:\n" + "\n".join(
36
+ [
37
+ f"- {row['artist_name']} 嘅 **[「{row['track_name']}」](https://open.spotify.com/track/{row['track_id']})**(相似度:{row['similarity']:.2f})"
38
+ for _, row in top_5.iterrows()
39
+ ]
40
+ )
41
+
42
+
43
+ def get_embedding(text):
44
+ return model.encode(text)
45
+
46
+
47
+ # Create Gradio application
48
+ import gradio as gr
49
+
50
+
51
+ async def create_demo():
52
+ example1 = """故事開始於一個悲傷的雨夜,主角站在街頭,淋雨中凝視著遠方。心裡充滿了對一段失去的愛情的痛苦和迷惘。回想起往事,他感到愛情並非他當初所想像的那麼美好,無法找到回到對方身邊的路,更不用說如何忘記過往。
53
+
54
+ 淚水在眼底打轉,他感到迷失,不知道該往哪裡去,只是心中不斷地呼喚著對方的名字,渴望重新找回失去的愛。他開始思考,是應該安靜地離開這段過往,還是勇敢地留下來,面對愛情的種種無奈。
55
+
56
+ 在迷惘中,他決定給自己一個機會。或許他應該離開,或者他應該在原地等待,等待對方明白他所付出的愛是永遠不會離開的。這段故事充滿了對愛情的掙扎、遺憾和無奈,卻也帶著一絲希望和堅持。結局如何,只有時間能給予答案。"""
57
+ example2 = "我唔搵唔返一啲嘢,搵唔返。"
58
+ example3 = "香港有國安法"
59
+ example4 = "雞!全部都係雞!"
60
+ description = """呢個 space 利用咗廣東話 Bert 語言模型,將歌詞轉換成向量,再用 cosine similarity 計算輸入嘅文字同 2394 首粵語歌唧歌詞之間嘅相似度,嚟畀出個歌名推介。
61
+
62
+ ⚠️ 要註意!呢個唔係關鍵字搜尋,而係用文字嘅意思去比對每首歌嘅歌詞內容,所以有時候會有啲奇怪嘅結果。
63
+ """
64
+ css = """
65
+ .output {
66
+ padding: 10px;
67
+ min-height: 200px;
68
+ }
69
+ """
70
+
71
+ demo = gr.Interface(
72
+ fn=find_songs,
73
+ css=css,
74
+ inputs=[
75
+ gr.Textbox(
76
+ label="求其打啲嘢去搵歌名",
77
+ lines=5,
78
+ placeholder="請輸入歌詞",
79
+ ),
80
+ ],
81
+ outputs=[
82
+ gr.Markdown(elem_classes="output"),
83
+ ],
84
+ examples=[example1, example2, example3, example4],
85
+ title="粵語流行歌相似度",
86
+ description=description,
87
+ analytics_enabled=False,
88
+ allow_flagging=False,
89
+ )
90
+
91
+ return demo
92
+
93
+
94
+ # Run the application
95
+ if __name__ == "__main__":
96
+ demo = asyncio.run(create_demo())
97
+ demo.launch()
emb.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42d86b95f9d4422c247ac916161fa1a86f8fb886a29414866365b6b5717f49b1
3
+ size 50336246
requirements.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ gradio==4.39.0
2
+ sentence_transformers==3.0.1