Update README.md
Browse files
README.md
CHANGED
|
@@ -44,8 +44,7 @@ Finch focuses on **messy and long-horizon finance & accounting workflows** that
|
|
| 44 |
|
| 45 |
> data entry/import, structuring/formatting, web search, cross-sheet/file retrieval, calculation, financial modeling, validation, translation, visualization, and reporting.
|
| 46 |
|
| 47 |
-
The workflows are derived from
|
| 48 |
-
|
| 49 |
- Enterprise **email threads** where collaborators naturally describe, discuss, and track workflows
|
| 50 |
- Large and messy **spreadsheets** with multimodal artifacts including text, tables, formulas, charts, pivots, images, etc
|
| 51 |
- Interlinked **PDFs and documents** that provide additional business context
|
|
@@ -53,7 +52,7 @@ The workflows are derived from **real-world enterprise workspaces** (primarily E
|
|
| 53 |
We adopt a three-step workflow labeling process:
|
| 54 |
|
| 55 |
1. **Inducing workflow types and instances** from real collaborative context in **enterprise email threads** ([Enron Corpus](https://en.wikipedia.org/wiki/Enron_Corpus): 500,000 emails from 150 executives and employees).
|
| 56 |
-
2. **Deriving concrete workflow instances** by analyzing changes across **spreadsheet versions** (15,000 versioned spreadsheet files from Enron and [EUSES](https://dl.acm.org/doi/10.1145/1082983.1083242)) and designing workflows based on high-quality artifacts from investment and securities companies, World Bank, Canadian/British government agencies, WideSearch, Dabstep, and more.
|
| 57 |
3. **Conducting meticulous expert annotation** of task instructions, input files, and reference outputs, involving hundreds of hours of expert work.
|
| 58 |
|
| 59 |
<img src="figs/annotation.png" width="1000" />
|
|
|
|
| 44 |
|
| 45 |
> data entry/import, structuring/formatting, web search, cross-sheet/file retrieval, calculation, financial modeling, validation, translation, visualization, and reporting.
|
| 46 |
|
| 47 |
+
The workflows are derived from real-world enterprise workspaces—primarily Enron and the EUSES Corpus—and further extended with more recent artifacts (e.g., from 2025) from investment and securities companies, the World Bank, Canadian and British government agencies, and other sources.
|
|
|
|
| 48 |
- Enterprise **email threads** where collaborators naturally describe, discuss, and track workflows
|
| 49 |
- Large and messy **spreadsheets** with multimodal artifacts including text, tables, formulas, charts, pivots, images, etc
|
| 50 |
- Interlinked **PDFs and documents** that provide additional business context
|
|
|
|
| 52 |
We adopt a three-step workflow labeling process:
|
| 53 |
|
| 54 |
1. **Inducing workflow types and instances** from real collaborative context in **enterprise email threads** ([Enron Corpus](https://en.wikipedia.org/wiki/Enron_Corpus): 500,000 emails from 150 executives and employees).
|
| 55 |
+
2. **Deriving concrete workflow instances** by analyzing changes across **spreadsheet versions** (15,000 versioned spreadsheet files from Enron and [EUSES](https://dl.acm.org/doi/10.1145/1082983.1083242)) and designing workflows based on high-quality recent artifacts from investment and securities companies, World Bank, Canadian/British government agencies, WideSearch, Dabstep, and more.
|
| 56 |
3. **Conducting meticulous expert annotation** of task instructions, input files, and reference outputs, involving hundreds of hours of expert work.
|
| 57 |
|
| 58 |
<img src="figs/annotation.png" width="1000" />
|