💩 dataset convert script

dataset convert script

app/config.py

@@ -7,7 +7,7 @@
 from dotenv import load_dotenv
 
 load_dotenv()
-session = boto3.Session()
+session = boto3.Session(profile_name="cs-broker")
 secret_manager = session.client(service_name="secretsmanager", region_name="ap-northeast-2")
 log = logging.getLogger("__main__")
 log.setLevel(logging.INFO)

app/tests/conftest.py

@@ -21,16 +21,16 @@ def content_model() -> PromptForClassification:
 
 
 @pytest.fixture(scope="session")
-def problem_df(path: str = "app/static/user_answer.csv") -> pd.DataFrame:
+def user_answer_df(path: str = "app/static/user_answer.csv") -> pd.DataFrame:
     return pd.read_csv(path)
 
 
 @pytest.fixture(scope="session")
-def problem_dict(keyword_model: SentenceTransformer, problem_df: pd.DataFrame) -> dict:
+def problem_dict(keyword_model: SentenceTransformer, user_answer_df) -> dict:
     problem_dict = {}
     keyword_id = 0
     # criterion parsing
-    for _, data in problem_df.iterrows():
+    for _, data in user_answer_df.iterrows():
         problem_id = data["problem_id"]
         if problem_id not in problem_dict:
             keyword_standards = []
@@ -49,9 +49,9 @@ def problem_dict(keyword_model: SentenceTransformer, problem_df: pd.DataFrame) -
 
 
 @pytest.fixture(scope="function")
-def random_keyword_data(problem_dict: dict, problem_df: pd.DataFrame) -> KeywordGradingRequest:
-    random_idx = random.randint(0, len(problem_df) - 1)
-    random_data = problem_df.iloc[random_idx]
+def random_keyword_data(problem_dict: dict, user_answer_df) -> KeywordGradingRequest:
+    random_idx = random.randint(0, len(user_answer_df) - 1)
+    random_data = user_answer_df.iloc[random_idx]
     problem_id = random_data["problem_id"]
     keyword_standards = problem_dict[problem_id].keyword_standards
     return KeywordGradingRequest(
@@ -60,9 +60,9 @@ def random_keyword_data(problem_dict: dict, problem_df: pd.DataFrame) -> Keyword
 
 
 @pytest.fixture(scope="function")
-def random_content_data(problem_df: pd.DataFrame) -> ContentGradingRequest:
-    random_idx = random.randint(0, len(problem_df) - 1)
-    random_series = problem_df.iloc[random_idx]
+def random_content_data(user_answer_df) -> ContentGradingRequest:
+    random_idx = random.randint(0, len(user_answer_df) - 1)
+    random_series = user_answer_df.iloc[random_idx]
     content_standards = []
     offset = random.randint(0, 10000)
     for i, criterion in enumerate(eval(random_series["scoring_criterion"])):
@@ -72,3 +72,14 @@ def random_content_data(problem_df: pd.DataFrame) -> ContentGradingRequest:
     return ContentGradingRequest(
         problem_id=random_idx, user_answer=random_series.user_answer, content_standards=content_standards
     )
+
+
+# from collections import defaultdict
+# @pytest.fixture(scope="session")
+# def get_problem_df(path: str = "/Users/minjaewon/workspace/AI-server/app/static/problem_info.csv"):
+#     df = pd.read_csv(path)
+#     problem_info = {}
+#     for row in df.iterrows():
+#
+#         if row.problem_id not in problem_info:
+#             problem_info[row.problem_id] = {"keyword_standards": }

app/tests/test_keyword.py

@@ -47,3 +47,10 @@ def test_keyword_predict_runnable_2(problem_dict: dict, random_keyword_data: Key
     problem_keyword_set = set(keyword.id for keyword in runnable.problem_dict[test_problem_id].keyword_standards)
     for correct_keyword in result.correct_keywords:
         assert correct_keyword.id in problem_keyword_set, "problem_dict에 맞지 않는 키워드를 이용해 예측하였습니다."
+
+
+def test_keyword_predict_runnable_3(problem_dict: dict, random_keyword_data: KeywordGradingRequest) -> None:
+    """
+    ','로 구분되어 있는 키워드 기준들을 함께 보면서 유사도 측정
+    """
+    print()

app/utils/utils.py

@@ -0,0 +1,103 @@
+import os
+
+import numpy as np
+import pandas as pd
+
+
+def update_problem_info(problem_info_path: str, user_answer_path: str):
+
+    problem_info_df = pd.read_csv(problem_info_path)
+
+    problem_info = {}
+    for i in range(len(problem_info_df)):
+        row = problem_info_df.iloc[i]
+        if row.problem_id not in problem_info:
+            problem_info[row.problem_id] = {"keyword_standards": [], "content_standards": []}
+        if row.content == np.nan:
+            row.content = "NULL"
+        if row.type == "KEYWORD":
+            problem_info[row.problem_id]["keyword_standards"].append(row.content)
+        elif row.type == "PROMPT":
+            problem_info[row.problem_id]["content_standards"].append(row.content)
+
+    user_answer_df = pd.read_csv(user_answer_path)
+
+    change_dict = {4: 449, 6: 453, 1: 454, 5: 450, 9: 447, 3: 451, 8: 448, 2: 446, 0: 452, 7: 443}
+    visit = set()
+    keyword_remove_dict = {}
+    content_remove_dict = {}
+    for i in range(len(user_answer_df)):
+        problem_id = change_dict[user_answer_df.iloc[i].problem_id]
+        converted_problem = problem_info[problem_id]
+        if problem_id not in visit:
+            print(user_answer_df.keyword_criterion[i])
+            print(converted_problem["keyword_standards"])
+            order = list(map(int, input().split()))
+            if len(eval(user_answer_df.keyword_criterion[i])) != len(converted_problem["keyword_standards"]):
+                print("삭제할 인덱스를 적어주세요")
+                keyword_remove_dict[user_answer_df.iloc[i].problem_id] = list(map(int, input().split()))
+
+            converted_problem["keyword_standards"] = [converted_problem["keyword_standards"][j - 1] for j in order]
+            print(user_answer_df.keyword_criterion[i])
+            print(converted_problem["keyword_standards"])
+            print("*" * 50)
+            print(user_answer_df.scoring_criterion[i])
+            print(converted_problem["content_standards"])
+            order = list(map(int, input().split()))
+            if len(eval(user_answer_df.scoring_criterion[i])) != len(converted_problem["content_standards"]):
+                print("삭제할 인덱스를 적어주세요")
+                content_remove_dict[user_answer_df.iloc[i].problem_id] = list(map(int, input().split()))
+
+            converted_problem["content_standards"] = [converted_problem["content_standards"][j - 1] for j in order]
+            print(user_answer_df.scoring_criterion[i])
+            print(converted_problem["content_standards"])
+            print("*" * 50)
+            visit.add(problem_id)
+        labeled_keyword_ids = []
+        correct_keyword_criterion = eval(user_answer_df.correct_keyword_criterion[i])
+        keyword_criterion = eval(user_answer_df.keyword_criterion[i])
+        if user_answer_df.iloc[i].problem_id in keyword_remove_dict:
+            for idx in keyword_remove_dict[user_answer_df.iloc[i].problem_id]:
+                keyword_criterion[idx - 1] = None
+        keyword_criterion = [value for value in keyword_criterion if value is not None]
+        keyword_criterion = [value if value != np.nan else "NULL" for value in keyword_criterion]
+        print(correct_keyword_criterion)
+        print(keyword_criterion)
+
+        for criterion in correct_keyword_criterion:
+            if criterion in keyword_criterion:
+                idx = keyword_criterion.index(criterion)
+                labeled_keyword_ids.append(idx)
+        print(labeled_keyword_ids)
+        print(converted_problem["keyword_standards"])
+        labeled_content_ids = []
+        correct_scoring_criterion = eval(user_answer_df.correct_scoring_criterion[i])
+        scoring_criterion = eval(user_answer_df.scoring_criterion[i])
+
+        if user_answer_df.iloc[i].problem_id in content_remove_dict:
+            for idx in content_remove_dict[user_answer_df.iloc[i].problem_id]:
+                scoring_criterion[idx - 1] = None
+        scoring_criterion = [value for value in scoring_criterion if value is not None]
+
+        print(scoring_criterion)
+        print(correct_scoring_criterion)
+        for criterion in correct_scoring_criterion:
+            if criterion in scoring_criterion:
+                idx = scoring_criterion.index(criterion)
+                labeled_content_ids.append(idx)
+        print(labeled_content_ids)
+        print(converted_problem["content_standards"])
+        print("*" * 50)
+        user_answer_df.correct_keyword_criterion[i] = labeled_keyword_ids
+        user_answer_df.correct_scoring_criterion[i] = labeled_content_ids
+        user_answer_df.keyword_criterion[i] = converted_problem["keyword_standards"]
+        user_answer_df.scoring_criterion[i] = converted_problem["content_standards"]
+
+    user_answer_df.to_csv(os.path.join(os.path.dirname(user_answer_path), "changed_user_answer.csv"))
+
+
+if __name__ == "__main__":
+    update_problem_info(
+        problem_info_path="/Users/minjaewon/workspace/AI-server/app/static/problem_info.csv",
+        user_answer_path="/Users/minjaewon/workspace/AI-server/app/static/user_answer.csv",
+    )