R3troR0b commited on
Commit
63604b6
1 Parent(s): d5b9808

Update README.md

Browse files

# Book Marker Dataset (book_marker_dataset)

## Overview
The `book_marker_dataset` splits entire books into manageable sections, using sequential markers to maintain continuity across book splits. Each book is divided into token-limited segments, with markers connecting the end of one section to the beginning of the next.

### Structure
Each entry consists of:
- **Book Segment**: A chunk of text from a book.
- **Sequential Marker**: The marker connecting this chunk to the next or previous one.

### Columns:
text: The column containing the current chunk of the book text.
label: The columns containing the sequential marker and the next chunk of the book

```json
{"text": "book_segment", "label": ["marker", "next_segment"]}
```

### Example:
```json
{
"book_segment": "America is again facing forward to the sea...",
"marker": "!1",
"next_segment": "!1 The ancient thrill of the wide salt spaces..."
}
```
### Training Task:
LLM Supervised Fine-Tuning (SFT): This dataset is useful for training long-context models on book-level text processing.

### Usage
This dataset is intended for tasks involving long-form document processing, such as book summarization or analysis. The sequential markers help models understand the connections between different parts of a book.

### Source
This dataset was generated by processing books and splitting them into token-limited chunks, ensuring continuity with sequential markers.

Files changed (1) hide show
  1. README.md +12 -3
README.md CHANGED
@@ -1,3 +1,12 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ task_categories:
4
+ - text-generation
5
+ language:
6
+ - en
7
+ tags:
8
+ - context
9
+ - sequential
10
+ - indices
11
+ pretty_name: Sequential Marking For Book Level
12
+ ---