Merge pull request #3 from dataforgoodfr/feat/multi_step_scrapping

Feat/multi step scrapping

.gitignore

@@ -4,7 +4,7 @@
 #     https://www.atlassian.com/git/tutorials/saving-changes/gitignore
 
 .env
-data/*
+data/.ipynb_checkpoints/
 
 
 # Node artifact files
@@ -53,7 +53,7 @@ Thumbs.db
 *.wmv
 
 *.pyc
-notebooks/.ipynb_checkpoints
+notebooks/**/.ipynb_checkpoints/
 .env
 .env
 .venv

climateguard/gdelt_scrapper.py

@@ -0,0 +1,76 @@
+from urllib.request import urlopen
+import pandas as pd
+import gdeltdoc as gdelt
+import functools
+import itertools
+from pathlib import Path
+
+class GDELTScrapper: 
+    THEMES_URL = "http://data.gdeltproject.org/api/v2/guides/LOOKUP-GKGTHEMES.TXT"
+
+    @functools.cached_property
+    def themes_df(self) -> pd.DataFrame:    
+        # Fetch the content using urllib
+        with urlopen(self.THEMES_URL) as response:
+            data = response.read().decode()
+
+        # Split the data into lines
+        lines = data.strip().split("\n")
+
+        # Split each line into key-value pairs
+        rows = [line.split("\t") for line in lines]
+
+        # Create a DataFrame from the rows
+        df = pd.DataFrame(rows, columns=['theme', 'count'])
+        df['count'] = df['count'].astype(int)
+
+        return df
+
+    def find_themes_related_to_keyword(self, keyword: str) -> list[str]: 
+        return self.themes_df[self.themes_df["theme"].str.contains(keyword, case=False)]["theme"].to_list()
+
+    def find_articles(self, themes: list[str], years: list[int]) -> pd.DataFrame: 
+        partial_articles_dfs = []
+
+        gd = gdelt.GdeltDoc()
+        for theme, year in itertools.product(themes, years): 
+            f = gdelt.Filters(
+                #keyword = "climate change",
+                start_date=f"{year}-01-01",
+                end_date=f"{year}-12-31", 
+                theme=theme, 
+                country="LG", # Latvia
+            )
+
+            partial_articles_df = gd.article_search(f)
+            print(f"{len(partial_articles_df)} articles found for theme {theme}, in {year}")
+            partial_articles_dfs.append(partial_articles_df)
+
+        articles_df = pd.concat(partial_articles_dfs)
+
+        articles_df = articles_df[articles_df["language"] == "Latvian"]
+        articles_df["seendate"] = pd.to_datetime(articles_df["seendate"])
+
+        print(f"Deleting {articles_df["url"].duplicated().sum()} duplicates")
+        articles_df = articles_df.drop_duplicates("url")
+        print(f"{len(articles_df)} unique articles found")
+        return articles_df
+
+
+# Usage example:
+if __name__ == "__main__":
+    scraper = GDELTScrapper()
+
+    # Find themes related to climate
+    themes = scraper.find_themes_related_to_keyword("CLIMATE")
+    print(f"Themes related to climate: {themes}")
+
+    # Find articles for these themes and year range
+    articles_df = scraper.find_articles(themes=themes, years=[2022, 2023, 2024])
+
+    # This can be used as input for NewsScraper
+    article_urls = articles_df["url"].to_list()
+
+    # Save dataframe to a csv file
+    file_path = Path(__file__).parent.parent / "data/latvian_article_links.csv"
+    articles_df.to_csv(file_path)