Update README.md
Browse files
README.md
CHANGED
@@ -62,9 +62,19 @@ Example (truncated):
|
|
62 |
}
|
63 |
```
|
64 |
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
|
|
|
|
|
|
|
|
|
68 |
Example (truncated):
|
69 |
|
70 |
```json
|
@@ -108,9 +118,13 @@ Each sample has five top-level fields:
|
|
108 |
| `text` | `str` | Full text with image placeholders |
|
109 |
| `image_info` | `list\<dict\>` | List of image metadata:<br>• `placeholder` (str)<br>• `url` (str)<br>• `original_width` / `original_height` (int)<br>• `exif` (dict, optional)<br>• `alt` (str, optional) |
|
110 |
|
|
|
|
|
|
|
|
|
111 |
## Size and Splits
|
112 |
|
113 |
-
MOMIJI contains about **56M** Japanese web documents. Because images are provided only as URLs (no binary data), the dataset is roughly **
|
114 |
|
115 |
## MOMIJI Statistics
|
116 |
|
|
|
62 |
}
|
63 |
```
|
64 |
|
65 |
+
Each sample has three top-level fields:
|
66 |
+
|
67 |
+
| Field | Type | Description |
|
68 |
+
|--------------|-------------------|---------------------------------------------------------------------------------------------------------------|
|
69 |
+
| `docId` | `str` | Unique ID derived from the source Common Crawl ID |
|
70 |
+
| `url` | `str` | Source web URL | |
|
71 |
+
| `image_info` | `list\<dict\>` | List of image metadata:<br>• `placeholder` (str)<br>• `url` (str)<br>• `original_width` / `original_height` (int)<br>• `exif` (dict, optional)<br>• `alt` (str, optional) |
|
72 |
|
73 |
|
74 |
+
MOMIJI contains the data fields listed above. Please use the **[momiji_generator](https://github.com/turingmotors/momiji_generator)** utility to populate the text and text_list fields. See the section below for an example record generated by momiji_generator, showing the populated `text` and `text_list` fields.
|
75 |
+
|
76 |
+
<details>
|
77 |
+
|
78 |
Example (truncated):
|
79 |
|
80 |
```json
|
|
|
118 |
| `text` | `str` | Full text with image placeholders |
|
119 |
| `image_info` | `list\<dict\>` | List of image metadata:<br>• `placeholder` (str)<br>• `url` (str)<br>• `original_width` / `original_height` (int)<br>• `exif` (dict, optional)<br>• `alt` (str, optional) |
|
120 |
|
121 |
+
</details>
|
122 |
+
|
123 |
+
|
124 |
+
|
125 |
## Size and Splits
|
126 |
|
127 |
+
MOMIJI contains about **56M** Japanese web documents. Because images are provided only as URLs (no binary data), the dataset is roughly **14.7 GB**.
|
128 |
|
129 |
## MOMIJI Statistics
|
130 |
|