Update README.md

README.md

@@ -14,7 +14,7 @@ size_categories:
 
 # Anime Manga Characters Dataset
 
-This dataset is a metafile containing information about anime and manga characters sourced from various fandom wikis. Each entry represents a character along with its associated metadata. The dataset has been deduplicated based on the `url` field to avoid redundancy, although a single character may still have multiple associated URLs.
+This dataset is a metafile containing information about 247,034 anime and manga characters sourced from various fandom wikis. Each entry represents a character along with its associated metadata. The dataset has been deduplicated based on the `url` field to avoid redundancy, although a single character may still have multiple associated URLs.
 
 ## Potential Applications
 
@@ -23,6 +23,37 @@ This dataset is a metafile containing information about anime and manga characte
 - **Image Captioning**: Pair images of characters with textual metadata for image captioning tasks.
 - **Text-to-Image Generation**: Generate detailed images of characters based on descriptive metadata.
 
+## Dataset Construction
+
+The dataset includes the following meta fields:
+
+- **`title`**: The name of the character, corresponding to the title of the fandom wiki page.
+- **`site_name`**: The title of the anime or manga, representing the source site name of the fandom wiki.
+- **`url`**: The URL of the character's wiki page.
+- **`description`**: A brief summary of the character's wiki page (truncated, not the full content).
+- **`image`**: The URL of a representative image associated with the character.
+
+The pseudo code for getting metadata from each fandom wiki html:
+
+```python
+def get_metadata(self, html):
+    soup = BeautifulSoup(html, 'lxml')
+    metadata = {"title": "", "site_name": "", "url": "", "description": "", "image": ""}
+    if meta := soup.find("meta", {"property": "og:site_name"}):
+        metadata["site_name"] = meta["content"]
+    if meta := soup.find("meta", {"property": "og:title"}):
+        metadata["title"] = meta["content"]
+    if meta := soup.find("meta", {"property": "og:url"}):
+        metadata["url"] = meta["content"]
+    if meta := soup.find("meta", {"property": "og:description"}):
+        metadata["description"] = meta["content"]
+    if meta := soup.find("meta", {"property": "og:image"}):
+        metadata["image"] = meta["content"]
+    return metadata
+```
+
+The dataset has been deduplicated based on the `url` field to avoid redundancy.
+
 ## Acknowledgments
 
 This dataset owes its existence to the passionate contributions of anime and manga fans worldwide who curate and maintain the fandom wikis.