> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.beatapi.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.beatapi.io/_mcp/server.

# How should I organize Bilibili comments into a course team's question table?

> Preserve video sources, excerpts, reply relationships and learning conditions, then hand off questions with actual collection scope.

**Build a table the course team can act on: video, viewer wording, confusing step, existing explanation and next check.** Read comments with `bilibili.app.fetch_video_comments`, expand ambiguous comments with `bilibili.web.fetch_comment_reply`, and keep the original comments on a separate sheet from your interpretation.

For example, after an editing-course update, viewers may ask about importing material, shortcuts or software versions. Select the relevant videos, read the actual comments and group the specific learning questions. Keep giveaway entries and general praise out of the question groups.

## 1. Save a relevant comment batch

Reuse a known BV ID, or establish the BV ID through the [BV/AV steps](/faq/bilibili-bv-comments). `bilibili.app.fetch_video_comments` returns time-ordered comments with `mode: 2`; the first page uses `next_offset: 1`:

```json
{
  "action": "bilibili.app.fetch_video_comments",
  "params": {"bv_id": "<confirmed_bv_id>", "mode": 2, "next_offset": 1}
}
```

Send the request to `POST https://api.beatapi.io/v1/social-data/call` with normal BeatAPI Bearer authentication. For popular discussion, run a separate request with `mode: 3`, and save each observation's time and source; see [ordering](/faq/bilibili-comment-order).

For each video, save the response, BV target, actual inputs, time and request ID. Read the comment list, identifier and text fields from the actual response. Because Bilibili comment actions pass through raw response payloads without fixed top-level paths, inspect the saved response instead of copying array paths from another service.

## 2. Which columns support the handoff?

Use two sheets. Put one comment per row in **Source comments** and one question per row in **Teaching backlog**. Reference the supporting comment rows from each backlog item so teammates can trace every backlog item back to the original wording and video.

| Column                                               | Purpose                                            |
| ---------------------------------------------------- | -------------------------------------------------- |
| Video BV, title and source                           | Identify discussion and chapter                    |
| Actual comment and parent identifiers when available | Match repeats and context                          |
| Exact excerpt and comment time when supplied         | Revisit wording and applicable period              |
| Software, device or step conditions                  | Explain the learning situation                     |
| Topic and classification reason                      | Explain grouping to the team                       |
| Existing explanation, open items and next step       | Arrange answers, documentation or course revisions |
| Collection time, ordering, page state and request ID | State material coverage and source                 |

Preserve the original wording and line breaks, and leave missing values blank. Put both sheets in one Excel workbook or export each sheet as a separate UTF-8 CSV. When importing a CSV into Excel, set the BV, comment ID and parent comment ID columns to Text to preserve long identifiers.

## 3. How do you clarify ambiguous comments?

A comment like “This key does nothing” may be about a shortcut, the operating system or window focus. Read the comment's parent and the relevant replies. The Web reply action `bilibili.web.fetch_comment_reply` takes the video's `bv_id` and the verified `rpid` of the parent comment:

```json
{
  "action": "bilibili.web.fetch_comment_reply",
  "params": {"bv_id": "<confirmed_bv_id>", "rpid": "<confirmed_root_comment_id>", "pn": 1}
}
```

`rpid` identifies the selected comment, not the creator `mid` or the video's AV ID. Save the reply request inputs separately from the App root-comment `next_offset`, and check the objects each action actually returns.

## 4. How does the table become a teaching backlog?

The table below illustrates how to convert viewer comments into teaching backlog items:

| Excerpt                            | Conditions to establish         | Follow-up                                         |
| ---------------------------------- | ------------------------------- | ------------------------------------------------- |
| “The new version has no such menu” | Version and demonstrated step   | Add version guidance or update the demonstration  |
| “Imported material has no sound”   | Material, track and audio setup | Retain supplied facts and list concrete questions |
| “The reply above solved it”        | Parent and actual solution      | Restore context and record a reusable answer      |

1. When a comment repeats across pages, keep one comment record and save every observation.
2. Keep root and reply relationships, and count supporting comments separately from distinct videos.
3. Count several follow-ups from one person as one learner, not several.
4. Attach excerpts and unresolved conditions to each theme.

## What does a backlog row look like in CSV?

Suppose viewers on two videos ask where to find the menu in a new release. Put each excerpt in the source sheet, then group both excerpts under one question in the backlog. In this example, `C01` and `C02` are your row references. Each source row still carries the actual video and comment identifiers.

```csv
question,source_rows,video_count,open_question,next_step
Where is the transition menu in the new version?,C01 / C02,2,Software version,Confirm the menu path and update the lesson notes
```

Before handing off the workbook, pick one backlog item and follow the linked row references back to the source comments and videos. Record which videos and comment pages you read so someone else can pick up where you left off.

For the assistant:

> Organize these course videos' comments into a teaching table with video, wording, conditions and open questions. Expand ambiguous discussion and preserve parent relationships. Attach representative excerpts, separating answerable questions, follow-up needs and course updates. State actual reading scope.

Budget the root-comment pages and selected reply pages with current [pricing](https://beatapi.io/pricing). Reuse saved results, and request more pages only for unresolved learning questions.