feat: llm-filter 两级过滤(规则层+五类分类+去重+长文本保护),工作流单线程 v4
- 规则层(不调 LLM):横线装饰/HTML 水印 token/URL/邮箱/单双 ASCII 字符直接删 - LLM 五类分类:garbage/overlay/noise 删,repeat/dialogue 留,未识别回退保留 - 按文本去重:相同文本只调一次 LLM(忽略空白/大小写),判定一致并省调用 - 长文本保护:≥min_keep_len 时 noise 不构成删除依据 - 真实任务 run_ac7f480a3ccb 验证:非规则误删 350→193(-45%),呻吟/对话保留 - ocr-subtitle 工作流 v4:pool 钉死单线程(1/1) - 回归夹具 testdata/ocr_srt_run_ac7f480a3ccb.srt(真实 1666 条 OCR 输出)
This commit is contained in:
@@ -2,10 +2,10 @@
|
||||
"id": "ocr-subtitle",
|
||||
"name": "字幕OCR提取",
|
||||
"description": "抽帧并 OCR 提取视频烧录字幕,经 LLM 过滤无意义内容后生成带时间轴的 SRT 基准数据。",
|
||||
"version": 3,
|
||||
"version": 4,
|
||||
"definition": {
|
||||
"name": "字幕OCR提取",
|
||||
"version": 3,
|
||||
"version": 4,
|
||||
"nodes": [
|
||||
{
|
||||
"id": "extract",
|
||||
@@ -27,7 +27,9 @@
|
||||
"id": "ocr",
|
||||
"node_type": "subtitle-ocr",
|
||||
"params": {
|
||||
"prompt": "提取图像中的文字,不要描述图片中的内容"
|
||||
"prompt": "提取图像中的文字,不要描述图片中的内容",
|
||||
"pool_min_workers": 1,
|
||||
"pool_max_workers": 1
|
||||
},
|
||||
"inputs": {
|
||||
"frames_manifest": "extract.frames_manifest"
|
||||
@@ -36,7 +38,10 @@
|
||||
{
|
||||
"id": "filter",
|
||||
"node_type": "llm-filter",
|
||||
"params": {},
|
||||
"params": {
|
||||
"pool_min_workers": 1,
|
||||
"pool_max_workers": 1
|
||||
},
|
||||
"inputs": {
|
||||
"srt_uri": "ocr.srt_uri"
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user