Initial Commit

.gitignore

@@ -0,0 +1 @@
+__pycache__/*

app.py

@@ -0,0 +1,121 @@
+from gradio import (
+    Blocks,
+    Tabs,
+    TabItem,
+    Textbox,
+    Markdown,
+    Button,
+    Group,
+    Label,
+    Row,
+    Examples,
+)
+from gensim.models import Word2Vec
+import numpy as np
+import re
+
+novel2vec: Word2Vec = Word2Vec.load("./novel2vec_skipgram_gensim4_100dim.model")
+
+
+def on_show_novel_vector_click(input_novel_title):
+    if input_novel_title not in novel2vec.wv:
+        return f"『{input_novel_title}』はデータベースにありません"
+
+    vector: np.ndarray = novel2vec.wv.get_vector(input_novel_title)
+    vector_text = " ".join(map(lambda x: "{:.3f}".format(x), vector))
+    return vector_text
+
+
+def create_show_novel_vector_tab():
+    with TabItem("小説ベクトルを見る"):
+        with Row():
+            submit_button: Button = Button("ベクトルを見る")
+            input_novel_title_textbox: Textbox = Textbox(label="小説の題名", max_lines=1)
+        output_markdown: Markdown = Markdown()
+        submit_button.click(
+            on_show_novel_vector_click,
+            inputs=input_novel_title_textbox,
+            outputs=output_markdown,
+        )
+
+
+def parse_novel_formula(novel_formula):
+    splitted: list = re.split(r"([\+\-])", novel_formula)
+    splitted: list = map(lambda x: x.replace(" ", ""), splitted)
+    splitted: list = filter(lambda x: not x == "", splitted)
+    positive: list = []
+    negative: list = []
+    adding_list: list = positive
+    for x in splitted:
+        if x == "+":
+            adding_list = positive
+        elif x == "-":
+            adding_list = negative
+        else:
+            adding_list.append(x)
+    return positive, negative
+
+
+def on_start_calc_click(input_novel_formula):
+    positive, negative = parse_novel_formula(input_novel_formula)
+    unknown_titles = list(filter(lambda x: x not in novel2vec.wv, positive + negative))
+    if unknown_titles:
+        error_message = "** 以下の題名がデータベースにありません **"
+        for x in unknown_titles:
+            error_message += f"\n * {x}"
+        return None, error_message
+
+    result = novel2vec.wv.most_similar_cosmul(
+        positive=positive, negative=negative, topn=20
+    )
+    label_data = {title: confidence for title, confidence in result}
+    return label_data, ""
+
+
+def create_calc_novels_tab():
+    with TabItem("小説の演算を行う"):
+        with Group():
+            with Row():
+                submit_button = Button("演算する", variant="primary")
+                input_novel_formula_textbox = Textbox(label="小説の題名または小説式")
+        with Group():
+            output_label = Label(label="演算結果")
+            output_markdown = Markdown("")
+        with Group():
+            Examples(
+                examples=["世界から猫が消えたなら", "君の膵臓を食べたい + 涼宮ハルヒの憂鬱 - 三日間の幸福"],
+                inputs=input_novel_formula_textbox,
+                outputs=[output_label, output_markdown],
+                fn=on_start_calc_click,
+            )
+            Markdown(
+                """
+            * 小説の題名を1つ入力→ **類似した小説を検索** できます。
+            * 小説の題名を足し算したり引き算する→ **小説を演算** できます。
+            """
+            )
+
+        submit_button.click(
+            on_start_calc_click,
+            inputs=input_novel_formula_textbox,
+            outputs=[output_label, output_markdown],
+        )
+
+
+def main():
+    with Blocks() as interface:
+        Markdown("""
+        ## Novel2Vec Console
+        Twitterハッシュタグ`名刺代わりの小説10選`をWord2Vecで学習させ、小説をベクトル化しました。
+        ツイートデータは[GINK03様が公開されているデータ](https://github.com/GINK03/novel_recommend/blob/master/var/shosetsu_dataset.csv)を用いました。
+        詳しくは[私の記事を参照](https://note.com/omiyayimo/n/n0301112dbcc7)ください。
+        """)
+        with Tabs():
+            create_calc_novels_tab()
+            create_show_novel_vector_tab()
+
+    interface.launch()
+
+
+if __name__ == "__main__":
+    main()

novel2vec_skipgram_gensim4_100dim.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4cc9693de2040a63333ec5220fe6ff39adceff24f239e8d5a8b7b6d2f647cdb
+size 3680495

requirements.txt

@@ -0,0 +1 @@
+gensim

tests.py

@@ -0,0 +1,4 @@
+def parse_novel_formula_test():
+    from app import parse_novel_formula
+    print(parse_novel_formula("-A+ B ++C-D+E"))
+parse_novel_formula_test()