Fix README: update schema to match actual format (Parquet, JSON-string behavioral fields, empty structs)
Browse files
README.md
CHANGED
|
@@ -54,7 +54,9 @@ This dataset has been anonymized:
|
|
| 54 |
|
| 55 |
## Data Format
|
| 56 |
|
| 57 |
-
The dataset is provided in
|
|
|
|
|
|
|
| 58 |
|
| 59 |
### Schema
|
| 60 |
|
|
@@ -63,7 +65,7 @@ The dataset is provided in JSONL format (JSON Lines), where each line is a compl
|
|
| 63 |
"task_id": 1,
|
| 64 |
"participant_id": "participant_000001",
|
| 65 |
"question": "What is ...",
|
| 66 |
-
"ground_truth":
|
| 67 |
"status": "completed",
|
| 68 |
"start_timestamp": "2024-01-15T10:30:00Z",
|
| 69 |
"end_timestamp": "2024-01-15T10:45:00Z",
|
|
@@ -81,13 +83,21 @@ The dataset is provided in JSONL format (JSON Lines), where each line is a compl
|
|
| 81 |
"familiarity": 2,
|
| 82 |
"difficulty": 1,
|
| 83 |
"first_search_query": "...",
|
| 84 |
-
"initial_guess": "..."
|
|
|
|
| 85 |
},
|
| 86 |
"post_task_annotation": {
|
| 87 |
"difficulty_actual": 3,
|
| 88 |
-
"aha_moment_type": "search_result"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
},
|
| 90 |
-
"cancel_annotation": null,
|
| 91 |
"trials": [
|
| 92 |
{
|
| 93 |
"trial_num": 1,
|
|
@@ -104,7 +114,7 @@ The dataset is provided in JSONL format (JSON Lines), where each line is a compl
|
|
| 104 |
{
|
| 105 |
"url": "https://...",
|
| 106 |
"text": "selected text",
|
| 107 |
-
"dom_position":
|
| 108 |
"relevance": 0.8,
|
| 109 |
"credibility": 0.9
|
| 110 |
}
|
|
@@ -115,9 +125,10 @@ The dataset is provided in JSONL format (JSON Lines), where each line is a compl
|
|
| 115 |
"url": "https://...",
|
| 116 |
"referrer": "https://...",
|
| 117 |
"dwell_time": 45,
|
| 118 |
-
"rrweb_record": [
|
| 119 |
-
"event_list": [
|
| 120 |
-
"mouse_moves": [
|
|
|
|
| 121 |
}
|
| 122 |
]
|
| 123 |
}
|
|
|
|
| 54 |
|
| 55 |
## Data Format
|
| 56 |
|
| 57 |
+
The dataset is provided in Parquet format. Each row is a complete task record.
|
| 58 |
+
|
| 59 |
+
Behavioral trace fields (`rrweb_record`, `event_list`, `mouse_moves`, `page_switch_record`) are stored as JSON strings due to their variable nested structure. All other fields are native types.
|
| 60 |
|
| 61 |
### Schema
|
| 62 |
|
|
|
|
| 65 |
"task_id": 1,
|
| 66 |
"participant_id": "participant_000001",
|
| 67 |
"question": "What is ...",
|
| 68 |
+
"ground_truth": "...",
|
| 69 |
"status": "completed",
|
| 70 |
"start_timestamp": "2024-01-15T10:30:00Z",
|
| 71 |
"end_timestamp": "2024-01-15T10:45:00Z",
|
|
|
|
| 83 |
"familiarity": 2,
|
| 84 |
"difficulty": 1,
|
| 85 |
"first_search_query": "...",
|
| 86 |
+
"initial_guess": "...",
|
| 87 |
+
"expected_source": ["search_engine"]
|
| 88 |
},
|
| 89 |
"post_task_annotation": {
|
| 90 |
"difficulty_actual": 3,
|
| 91 |
+
"aha_moment_type": "search_result",
|
| 92 |
+
"strategy_shift": ["..."],
|
| 93 |
+
"strategy_shift_other": "",
|
| 94 |
+
"unhelpful_paths": ["..."]
|
| 95 |
+
},
|
| 96 |
+
"cancel_annotation": {
|
| 97 |
+
"category": [],
|
| 98 |
+
"reason": "",
|
| 99 |
+
"missing_resource": ""
|
| 100 |
},
|
|
|
|
| 101 |
"trials": [
|
| 102 |
{
|
| 103 |
"trial_num": 1,
|
|
|
|
| 114 |
{
|
| 115 |
"url": "https://...",
|
| 116 |
"text": "selected text",
|
| 117 |
+
"dom_position": "CSS selector",
|
| 118 |
"relevance": 0.8,
|
| 119 |
"credibility": 0.9
|
| 120 |
}
|
|
|
|
| 125 |
"url": "https://...",
|
| 126 |
"referrer": "https://...",
|
| 127 |
"dwell_time": 45,
|
| 128 |
+
"rrweb_record": "[{...}]",
|
| 129 |
+
"event_list": "[{...}]",
|
| 130 |
+
"mouse_moves": "[{...}]",
|
| 131 |
+
"page_switch_record": "[{...}]"
|
| 132 |
}
|
| 133 |
]
|
| 134 |
}
|