From 0644bfa5a3f9fcf8b4e9c818e3e95ec842fb6c00 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E2=80=9Cweifeng2333=E2=80=9D?= <“2715673327@qq.com”> Date: Wed, 13 Nov 2024 00:24:38 +0800 Subject: [PATCH] update readme --- README.md | 6 +-- app/core/alingner_wag1.py | 110 -------------------------------------- 2 files changed, 3 insertions(+), 113 deletions(-) delete mode 100644 app/core/alingner_wag1.py diff --git a/README.md b/README.md index 22d0edd..2df8ee6 100644 --- a/README.md +++ b/README.md @@ -1,14 +1,14 @@
VideoCaptioner Logo +

卡卡字幕助手

VideoCaptioner

一款基于大语言模型(LLM)的智能视频字幕处理助手，支持字幕生成、断句、优化、翻译全流程处理

- - 简体中文 / [English](./docs/README_EN.md) +

简体中文 / [English](./docs/README_EN.md)

## 📖 项目介绍 -VideoCaptioner 是一款功能强大的视频字幕配制软件。操作简单且无需高配置，利用大语言模型进行字幕智能断句、校正、优化、翻译，一键为视频配上效果惊艳的字幕。 +卡卡字幕助手（VideoCaptioner）是一款功能强大的视频字幕配制软件。操作简单且无需高配置，利用大语言模型进行字幕智能断句、校正、优化、翻译，一键为视频配上效果惊艳的字幕。 - 🎯 无需GPU即可使用强大的语音识别引擎，自动生成精准字幕 - ✂️ 基于 LLM 的智能分割与断句，字幕阅读更自然流畅 diff --git a/app/core/alingner_wag1.py b/app/core/alingner_wag1.py deleted file mode 100644 index 041f0fa..0000000 --- a/app/core/alingner_wag1.py +++ /dev/null @@ -1,110 +0,0 @@ -from rapidfuzz import fuzz - - -def align_texts_sequence_alignment(text1, text2, gap_penalty=100): - """ - 使用序列比对算法（如 Needleman-Wunsch）对两个文本列表进行对齐， - 不改变原始列表的内容，只是在对齐过程中可能插入空字符串。 - - Args: - text1 (list): 源文本列表，保持不变。 - text2 (list): 目标文本列表，保持不变。 - gap_penalty (int): 插入或删除操作的固定惩罚分数。 - - Returns: - tuple: (aligned_source, aligned_target) - - aligned_source (list): 对齐后的源文本列表。 - - aligned_target (list): 对齐后的目标文本列表。 - """ - n = len(text1) - m = len(text2) - - # 初始化得分矩阵和方向矩阵 - score = [[0] * (m + 1) for _ in range(n + 1)] - direction = [[None] * (m + 1) for _ in range(n + 1)] - - # 初始化第一列和第一行 - for i in range(1, n + 1): - score[i][0] = score[i - 1][0] - gap_penalty - direction[i][0] = 'up' # 从上方来，表示插入空的目标句子 - for j in range(1, m + 1): - score[0][j] = score[0][j - 1] - gap_penalty - direction[0][j] = 'left' # 从左方来，表示插入空的源句子 - - # 填充得分矩阵和方向矩阵 - for i in range(1, n + 1): - for j in range(1, m + 1): - # 计算匹配得分 - sim = fuzz.ratio(text1[i - 1], text2[j - 1]) - match_score = score[i - 1][j - 1] + sim # 相似度越高，得分越高 - - # 计算插入和删除得分 - delete_score = score[i - 1][j] - gap_penalty # 删除 text1 中的句子 - insert_score = score[i][j - 1] - gap_penalty # 插入 text2 中的句子 - - # 选择得分最高的操作 - max_score = max(match_score, delete_score, insert_score) - score[i][j] = max_score - - # 记录方向 - if max_score == match_score: - direction[i][j] = 'diag' # 来自左上方，表示匹配或替换 - elif max_score == delete_score: - direction[i][j] = 'up' # 来自上方，表示删除（text2 中插入空行） - else: - direction[i][j] = 'left' # 来自左方，表示插入（text1 中插入空行） - - # 回溯得到对齐结果 - aligned_source = [] - aligned_target = [] - i, j = n, m - while i > 0 or j > 0: - dir = direction[i][j] - if dir == 'diag': - aligned_source.insert(0, text1[i - 1]) - aligned_target.insert(0, text2[j - 1]) - i -= 1 - j -= 1 - elif dir == 'up': - aligned_source.insert(0, text1[i - 1]) - aligned_target.insert(0, '') # 插入空的目标句子 - i -= 1 - elif dir == 'left': - aligned_source.insert(0, '') # 插入空的源句子 - aligned_target.insert(0, text2[j - 1]) - j -= 1 - else: - break # 回溯结束 - - return aligned_source, aligned_target - - -if __name__ == '__main__': - text1 = ['yep human hair be about that thick', "yep and that's a really really tiny LED", - 'uvleds could be used to sterilize surfaces', 'like in hospitals or kitchens', 'just flick on the UV', - 'lights and pathogens would be dead in seconds', 'copy 19 or you know', - "UV LED companies stop pressing like it's kind of better because", - "everything's very good in these UV LEDs", 'you can start at all the covid 19', - 'for anything there we use aluminium gardenizer them', 'for UB we use aluminium gardenizer', - 'okay the Bam Jap is much bigger', "do you think this is what's coming", "it's okay to work", - 'but the problem the cost costs are too high changes', 'this is not thin passing', 'the cost is very high', - 'okay if the infinishing program', 'on a shifty pass closely is almost comparable'] - text2 = ['Yep, human hair is about that thick', - "Yep, and that's a really tiny LEDUV LEDs could be used to sterilize surface", - 'Like in hospitals or kitchens', 'Just flick on the UV lights and pathogens would be dead in seconds', - 'COVID-19 or you know', "UV LED companies are improving, it's kind of better because", - "everything's very good in these UV LEDs", 'You can start with all the COVID-19 precautions', - 'For everything, we use aluminum ganizers', 'For UV, we use aluminum ganizers', - 'Okay, the bigger one is much better', "Do you think this is what's coming?", "It's okay to work", - 'But the problem is the costs are too high', 'This is not a thin pass', 'The costs are very high', - 'Okay, if the finishing program', 'on a shifty pass closely is almost comparable'] - - aligned_source, aligned_target = align_texts_sequence_alignment(text1, text2) - - for idx, (s, t) in enumerate(zip(aligned_source, aligned_target)): - print(f"行 {idx + 1}:") - print(f"文本1: {s}") - print(f"文本2: {t}") - sim = fuzz.ratio(s, t) if s and t else 0 - print(f"相似度: {sim:.2f}") - print('----')