Sign in to view source links and access this dataset
Description
92,421 lines of Chinese game dialogue and narration extracted from Honkai Impact 3rd gameplay videos, covering story chapters from 'Main Story 1: Dusk, Maiden, Warship' to 'Main Story Part 2 Chapter 03 Interlude: A Sleepwalker's Pain'. The dataset was created by author mrzjy using an AI pipeline involving video frame extraction, OCR, and visual language model processing, and was last updated on December 17, 2024.
Use Cases
Train Chinese dialogue generation models based on character-labeled conversations.
Analyze narrative structure and character interactions in a long-form game story.
Benchmark Chinese OCR and visual-language model performance on game UI text extraction.
Study the evolution of storytelling and dialogue across sequential game chapters.
Strengths
Contains 92,421 structured lines of dialogue and narration.
Covers a defined narrative sequence from early to late game chapters.
Describes a multi-stage AI pipeline (video download, frame extraction, OCR, VLM parsing) used for creation.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Data may reflect bias inherent to the specific video sources and AI extraction pipeline used.
Provenance
Source
Extracted from Honkai Impact 3rd gameplay videos on Bilibili.
Collection Method
AI pipeline: video download (BBDown), frame segmentation, OCR (Paddle-OCR), structured parsing (MiniCPM-V-2_6), rule-based post-processing.
Time Range
Covers game story chapters from 'Main Story 1' to 'Main Story Part 2 Chapter 03 Interlude'.
Freshness
Last updated 2024-12-17 01:16:56; freshness should be verified.
License is unknown; terms of use must be verified before application.