Dannong Wang commited on
Commit
d642557
·
1 Parent(s): da76932

add dataset

Browse files
README.md ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ configs:
3
+ - config_name: tags
4
+ data_files:
5
+ - split: train
6
+ path: "xbrl_tags_train.csv"
7
+ - split: test
8
+ path: "xbrl_tags_test.csv"
9
+ - config_name: value
10
+ data_files:
11
+ - split: train
12
+ path: "value_train.csv"
13
+ - split: test
14
+ path: "value_test.csv"
15
+ - config_name: formula
16
+ data_files:
17
+ - split: train
18
+ path: "formula_formatted_with_tags_train.csv"
19
+ - split: test
20
+ path: "formula_formatted_with_tags_test.csv"
21
+ - config_name: formula_calculations
22
+ data_files:
23
+ - split: train
24
+ path: "formula_calculation_train.csv"
25
+ - split: test
26
+ path: "formula_calculation_test.csv"
27
+ ---
28
+
29
+ # XBRL Extraction Dataset
formula_calculation_test.csv ADDED
The diff for this file is too large to render. See raw diff
 
formula_calculation_train.csv ADDED
The diff for this file is too large to render. See raw diff
 
formula_formatted_with_tags_test.csv ADDED
The diff for this file is too large to render. See raw diff
 
formula_formatted_with_tags_train.csv ADDED
The diff for this file is too large to render. See raw diff
 
generate_xbrl_extract_hf_split.py ADDED
@@ -0,0 +1,119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import xml.etree.ElementTree as ET
2
+ import re
3
+ import json
4
+ from typing import List, Dict
5
+ from tqdm import tqdm
6
+ import random
7
+ import os.path
8
+ from huggingface_hub import HfApi
9
+ import csv
10
+
11
+ random.seed(42)
12
+ import subprocess
13
+
14
+ example_qa_dict = {
15
+ "xbrl_tags": {
16
+ "q": "What is the US GAAP XBRL tag for Cash and Cash Equivalents as reported by Example Company Inc for the Fiscal Year ending in FY 2022",
17
+ "a": "us-gaap:AnExampleTagName"
18
+ },
19
+ "value": {
20
+ "q": "What is the value of Exapmle company's income for the Fiscal year ending in FY 2020?",
21
+ "a": "80000000"
22
+ },
23
+ "formula_calculation": {
24
+ "q": "Can you provide the formula for Operating Profit Margin from Example Corp for the Fiscal Year ending in FY 2022?",
25
+ "a": "(50000000 / 3590000000) * 100"
26
+ },
27
+ "formula_formatted_with_tags": {
28
+ "q": "What is the formula for the Gross Profit Margin of Example Inc, formatted with the relevant US GAAP XBRL tags",
29
+ "a": "us-gaap:ExampleTag / us-gaap:AnotherExampleTag) * 100"
30
+ }
31
+ }
32
+
33
+
34
+ def get_xbrl_dataset(data: List[Dict], cat):
35
+ """
36
+ Saves entries with matching category1 or category2 in the format for fine-tuning.
37
+
38
+ Args:
39
+ data (List[Dict]): The input JSON data.
40
+ category (str): The category name to match.
41
+ """
42
+ results = []
43
+ for entry in data:
44
+ question = entry["query"]
45
+ question = re.sub(r"\(.*?\)", "", question)
46
+ context_ids = entry["contextID"]
47
+
48
+ # if not os.path.isfile('train/DowJones30/' + doc_path):
49
+ # print(f"missing file {doc_path}")
50
+ # continue
51
+
52
+ example_qa = f"Example question: {example_qa_dict[cat]['q']}\nExample answer: {example_qa_dict[cat]['a']}"
53
+ output = entry["raw_answer"]
54
+
55
+ if cat == 'formula_calculation':
56
+ question += " Answer with a formula substituted with values. "
57
+ output = entry["value_formula_answer"]
58
+
59
+ output = str(output)
60
+ instruction = (f"You are a knowledgeable XBRL assistant. Your task is to analyze the XBRL context and provide an"
61
+ f" accurate and very concise answer to the question, The example question can help you to learn the "
62
+ f"answer format. DO NOT output xml, code, explanation or create new question. \n{example_qa}\n")
63
+
64
+ input = f"Question: {question}\nAnswer:"
65
+ year = int(entry["{fiscal year/quarter}"].replace("FY ", ""))
66
+
67
+ # context_xml = add_xml(instruction + input, doc_path}, context_ids[0])
68
+ # if len(context_xml) > 24000:
69
+ # continue
70
+
71
+ # print(entry["answer"])
72
+ # entry["doc_path"], entry["answer"], entry["contextID"][0]
73
+ results.append({
74
+ "instruction": instruction,
75
+ "input": input,
76
+ "output": output,
77
+ "year": year,
78
+ "company": entry["ticker"],
79
+ "doc_path": entry['doc_path'],
80
+ "context_id": context_ids[0],
81
+ })
82
+
83
+ print("final length", len(results))
84
+ return results
85
+
86
+
87
+ def gen_xbrl(cat):
88
+ with open("xbrl_bench_34020.json", "r") as f:
89
+ data = json.load(f)
90
+ filtered_data = [entry for entry in data if entry['category1'] == cat or entry['category2'] == cat]
91
+
92
+ all_doc_path = list(set([entry['doc_path'] for entry in filtered_data]))
93
+ print(f"Total data size for {cat}: {len(filtered_data)}, total number of filings {len(all_doc_path)}")
94
+ random.shuffle(filtered_data)
95
+
96
+ dataset = get_xbrl_dataset(filtered_data, cat)
97
+
98
+ test_data = [x for x in dataset if x['year'] == 2023]
99
+ train_data = [x for x in dataset if x['year'] != 2023]
100
+
101
+ print(f"train size: {len(train_data)}, test size: {len(test_data)}\n")
102
+
103
+ with open(f"{cat}_test.csv", "w", newline="") as f:
104
+ w = csv.DictWriter(f, test_data[0].keys(), quoting=csv.QUOTE_ALL)
105
+ w.writeheader()
106
+ w.writerows(test_data)
107
+ with open(f"{cat}_train.csv", "w", newline="") as f:
108
+ w = csv.DictWriter(f, train_data[0].keys(), quoting=csv.QUOTE_ALL)
109
+ w.writeheader()
110
+ w.writerows(train_data)
111
+
112
+ return train_data, test_data
113
+
114
+
115
+ if __name__ == '__main__':
116
+ tags_train, tags_test = gen_xbrl("xbrl_tags")
117
+ value_train, value_test = gen_xbrl("value")
118
+ formula_train, formula_test = gen_xbrl("formula_formatted_with_tags")
119
+ formula_calc_train, formula_calc_test = gen_xbrl("formula_calculation")
value_test.csv ADDED
The diff for this file is too large to render. See raw diff
 
value_train.csv ADDED
The diff for this file is too large to render. See raw diff
 
xbrl_tags_test.csv ADDED
The diff for this file is too large to render. See raw diff
 
xbrl_tags_train.csv ADDED
The diff for this file is too large to render. See raw diff