Update README.md
Browse files
README.md
CHANGED
@@ -7,4 +7,29 @@ language:
|
|
7 |
pretty_name: MT-Mind2Web
|
8 |
tags:
|
9 |
- web navigation
|
10 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
pretty_name: MT-Mind2Web
|
8 |
tags:
|
9 |
- web navigation
|
10 |
+
---
|
11 |
+
# MT-Mind2Web Dataset
|
12 |
+
## Dataset Structure
|
13 |
+
- "task_id" (str): unique id for each task
|
14 |
+
- "website" (str): website name
|
15 |
+
- "domain" (str): website domain
|
16 |
+
- "subdomain" (str): website subdomain
|
17 |
+
- "turns" (list[dict]): list of subtasks
|
18 |
+
- "annotation_id" (str): unique id for each subtask
|
19 |
+
- "confirmed_task" (str): subtask description
|
20 |
+
- "action_reprs" (list[str]): human readable string representation of the action sequence
|
21 |
+
- "actions" (list[dict]): list of actions (steps) to complete the subtask
|
22 |
+
- "action_uid" (str): unique id for each action (step)
|
23 |
+
- "raw_html" (str): raw html of the page before the action is performed
|
24 |
+
- "cleaned_html" (str): cleaned html of the page before the action is performed
|
25 |
+
- "operation" (dict): operation to perform
|
26 |
+
- "op" (str): operation type, one of CLICK, TYPE, SELECT
|
27 |
+
- "original_op" (str): original operation type, contain additional HOVER and ENTER that are mapped to CLICK, not used
|
28 |
+
- "value" (str): optional value for the operation, e.g., text to type, option to select
|
29 |
+
- "pos_candidates" (list[dict]): ground truth elements. Here we only include positive elements that exist in "cleaned_html" after our preprocessing, so "pos_candidates" might be empty. The original labeled element can always be found in the "raw_html".
|
30 |
+
- "tag" (str): tag of the element
|
31 |
+
- "is_original_target" (bool): whether the element is the original target labeled by the annotator
|
32 |
+
- "is_top_level_target" (bool): whether the element is a top level target find by our algorithm. please see the paper for more details.
|
33 |
+
- "backend_node_id" (str): unique id for the element
|
34 |
+
- "attributes" (str): serialized attributes of the element, use `json.loads` to convert back to dict
|
35 |
+
- "neg_candidates" (list[dict]): other candidate elements in the page after preprocessing, has similar structure as "pos_candidates"
|