sec-filing-evidence-extractor

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

SEC Filing Evidence Extractor

SEC文件证据提取器

Extract evidence before drawing conclusions. Favor traceable, section-anchored facts over narrative summaries.

在得出结论前先提取证据。优先选择可追溯、锚定到具体章节的事实，而非叙述性摘要。

Workflow

工作流程

Identify filing set and periods.
Prefer at least 8 quarters plus latest annual filing when available.
Extract facts into a normalized evidence table.
Keep every row tied to a filing, period, and location anchor.
Flag uncertain interpretation explicitly instead of forcing classification.

确定文件集和期间。
若有可用数据，优先选择至少8个季度的文件加上最新的年度文件。
将事实提取到标准化的证据表格中。
确保每一行都关联到对应的文件、期间和位置锚点。
明确标记存在疑问的解读，而非强行进行分类。

Evidence Table Schema

证据表格 Schema

Use this schema in Markdown or CSV.

```
evidence_id
```
: stable id (for example
```
EV-001
```
)
```
filing
```
: form + date (for example
```
10-Q 2025-11-03
```
)
```
period
```
: fiscal period covered
```
location
```
: section/footnote anchor
```
topic
```
: revenue, receivables, reserves, CFFO, non-GAAP, acquisitions, etc.
```
fact
```
: concise paraphrase of disclosure
```
numbers
```
: raw numbers with units and sign
```
trend
```
: increase/decrease/flat/volatile/one-off
```
initial_tags
```
: optional candidate tags (
```
EM4
```
,
```
CF2
```
,
```
AA1
```
, etc.)
```
confidence
```
:
```
high|medium|low
```
for extraction accuracy only

使用Markdown或CSV格式的以下Schema。

```
evidence_id
```
: 稳定的ID（例如
```
EV-001
```
）
```
filing
```
: 表格类型+日期（例如
```
10-Q 2025-11-03
```
）
```
period
```
: 涵盖的会计期间
```
location
```
: 章节/附注锚点
```
topic
```
: 收入、应收账款、准备金、经营活动现金流（CFFO）、非GAAP指标、收购等
```
fact
```
: 披露内容的简洁转述
```
numbers
```
: 带单位和符号的原始数值
```
trend
```
: 上升/下降/持平/波动/一次性
```
initial_tags
```
: 可选的候选标签（
```
EM4
```
、
```
CF2
```
、
```
AA1
```
等）
```
confidence
```
: 仅针对提取准确性的
```
高|中|低
```

Extraction Priorities

提取优先级

Prioritize sections most likely to contain manipulative reporting clues.

Revenue recognition policy and contract assets/liabilities notes
Accounts receivable, allowance, DSO-related disclosures
Capitalized costs, deferred costs, useful lives, impairment
Restructuring, contingencies, reserves, and releases
Statement of cash flows plus supplemental cash disclosures
Non-GAAP reconciliations and KPI definition changes
Acquisition accounting, purchase price allocation, pro forma metrics

优先处理最可能包含操纵性报告线索的章节。

收入确认政策和合同资产/负债附注
应收账款、坏账准备、应收账款周转天数（DSO）相关披露
资本化成本、递延成本、使用年限、减值
重组、或有事项、准备金及释放
现金流量表加上补充现金披露
非GAAP指标调节和关键绩效指标（KPI）定义变更
收购会计、购买价格分配、备考指标

Quality Rules

质量规则

Separate disclosure facts from interpretation.
Keep quote snippets short; prefer paraphrase with exact numbers.
Normalize sign conventions for cash flow and expenses.
Capture definition changes in KPIs as separate evidence rows.
If wording is ambiguous, add a
```
low
```
extraction confidence row and continue.

将披露事实与解读分开。
引用片段要简短；优先使用带有精确数值的转述。
统一现金流和费用的符号约定。
将KPI的定义变更作为单独的证据行记录。
若表述模糊，添加一行提取置信度为
```
低
```
的记录，然后继续处理。

Few-Shot Example

少样本示例

The following abbreviated evidence table shows correctly extracted rows from a hypothetical 10-Q. Use this as an anchor for format, granularity, and tone.

evidence_id	filing	period	location	topic	fact	numbers	trend	initial_tags	confidence
EV-001	10-Q 2025-11-03	Q3 FY2025	Note 3 – Revenue Recognition	revenue	Company changed from point-in-time to over-time recognition for professional services contracts starting Q3	n/a	one-off	EM1	high
EV-002	10-Q 2025-11-03	Q3 FY2025	Balance Sheet, Note 5	receivables	Gross AR rose 34% YoY while revenue grew 12%; allowance for doubtful accounts held flat at $4.2M (1.8% of AR vs 2.6% prior year)	AR $233M vs $174M; allowance $4.2M vs $4.5M	increase	EM1, EM5	high
EV-003	10-Q 2025-11-03	Q3 FY2025	Cash Flow Statement	CFFO	CFO declined to $18M from $47M YoY; $22M increase in AR was the largest working-capital drag	CFO $18M vs $47M; AR drag -$22M	decrease	CF3	medium

Key points illustrated:

Each row captures one disclosure fact, not a conclusion.
Numbers are exact with units and comparative context.
Tags are candidates only; classification happens downstream.
The revenue-recognition policy change (EV-001) and the AR buildup (EV-002) are separate rows despite being related, preserving granularity.

以下简化的证据表格展示了从假设的10-Q文件中正确提取的行。以此作为格式、粒度和语气的参考。

evidence_id	filing	period	location	topic	fact	numbers	trend	initial_tags	confidence
EV-001	10-Q 2025-11-03	Q3 FY2025	附注3——收入确认	收入	公司从Q3开始将专业服务合同的收入确认时点从时点法变更为时段法	无	一次性	EM1	高
EV-002	10-Q 2025-11-03	Q3 FY2025	资产负债表，附注5	应收账款	应收账款总额同比增长34%，而收入仅增长12%；坏账准备维持在420万美元不变（占应收账款的1.8%，上年同期为2.6%）	应收账款 2.33亿美元 vs 1.74亿美元；坏账准备 420万美元 vs 450万美元	上升	EM1, EM5	高
EV-003	10-Q 2025-11-03	Q3 FY2025	现金流量表	经营活动现金流（CFFO）	经营活动现金流同比从4700万美元降至1800万美元；应收账款增加2200万美元是营运资金的最大拖累	经营活动现金流 1800万美元 vs 4700万美元；应收账款拖累 -2200万美元	下降	CF3	中

示例说明要点：

每一行记录一个披露事实，而非结论。
数值精确，带有单位和对比背景。
标签仅为候选；分类在后续步骤进行。
收入确认政策变更（EV-001）和应收账款增加（EV-002）虽相关，但分为单独行，以保留粒度。

Output Handoff

输出交付

Return:

Evidence table
Missing-data list (what was unavailable)
Ready-for-classification package keyed by
```
evidence_id
```

For section anchors and filing hotspots, read

references/sec-filing-hotspots.md

证据表格
缺失数据列表（无法获取的内容）
按
```
evidence_id
```
索引的待分类数据包

如需了解章节锚点和文件重点区域，请阅读

references/sec-filing-hotspots.md

。