メニュー
Process filler word annotations to generate video edit lists. Use when working with timestamp annotations for removing speech disfluencies (um, uh, like, you know) from audio/video content.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/benchflow-ai/skillsbench --skill filler-word-processingコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SkillsBench contribution workflow. Use when: (1) Creating benchmark tasks, (2) Understanding repo structure, (3) Preparing PRs for task submission.
SkillsBench task authoring — walk a contributor from idea to submission-ready task following CONTRIBUTING.md and the task-implementation rubric. Use when the user wants to create a new SkillsBench task, scaffold a task from an existing workflow (notebook, Excel workbook, document, dataset), convert a prompt or a benchmark item into a SkillsBench task, write skills for a task, or prepare a SkillsBench PR. Pairs with `task-review` (run that as a self-check before submitting).
SkillsBench task PR review — classifies the task track (standard / research / multimodal), runs static policy checks against the track-specific rubric, benchmarks the task across oracle plus Claude and Codex (with and without skills), audits trajectories for cheating and skill invocation, and produces a `pr-N-task-timestamp-run.txt` review report alongside a `prN.zip` bundle of trajectories. Use when reviewing a SkillsBench task PR (by number, branch, or local task path), when the user asks to review a task, run benchmarks on a PR, audit a submission, classify a task as research or multimodal track, or prepare a comment to post on a SkillsBench PR.
SOC 職業分類に基づく
SKILL.md を表示中
| name | filler-word-processing |
| description | Process filler word annotations to generate video edit lists. Use when working with timestamp annotations for removing speech disfluencies (um, uh, like, you know) from audio/video content. |
Typical annotation JSON structure:
[
{"word": "um", "timestamp": 12.5},
{"word": "like", "timestamp": 25.3},
{"word": "you know", "timestamp": 45.8}
]
Each filler word annotation marks when the word starts. To remove it, use word-specific durations since different fillers have different lengths:
json
WORD_DURATIONS = {
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
: ,
}
DEFAULT_DURATION =
():
(annotations_file) f:
annotations = json.load(f)
segments = []
ann annotations:
word = ann.get(, ).lower().strip()
timestamp = ann[]
word_duration = WORD_DURATIONS.get(word, DEFAULT_DURATION)
start = (, timestamp - buffer)
end = timestamp + word_duration
segments.append((start, end))
segments
When filler words are close together, merge their cut segments:
def merge_overlapping_segments(segments, min_gap=0.1):
"""
Merge segments that overlap or are very close together.
Args:
segments: List of (start, end) tuples
min_gap: Minimum gap to keep segments separate
Returns:
Merged list of segments
"""
if not segments:
return []
# Sort by start time
sorted_segs = sorted(segments)
merged = [sorted_segs[0]]
for start, end in sorted_segs[1:]:
prev_start, prev_end = merged[-1]
# If this segment overlaps or is very close to previous
if start <= prev_end + min_gap:
# Extend the previous segment
merged[-1] = (prev_start, max(prev_end, end))
else:
merged.append((start, end))
return merged
def process_filler_annotations(annotations_file, word_duration=0.4):
"""Full pipeline: load annotations -> create segments -> merge overlaps"""
# Load and create initial segments
segments = annotations_to_segments(annotations_file, word_duration)
# Merge overlapping cuts
merged = merge_overlapping_segments(segments)
return merged
| Parameter | Typical Value | Notes |
|---|---|---|
| word_duration | varies | Short fillers (um, uh) ~0.25-0.3s, single words (like, yeah) ~0.3-0.4s, phrases (you know, i mean) ~0.5-0.6s |
| buffer | 0.05s | Small buffer captures word onset |
| min_gap | 0.1s | Prevents micro-segments between close fillers |
| Category | Words | Duration |
|---|---|---|
| Quick hesitations | uh, um | 0.3-0.4s |
| Sustained hums (drawn out while thinking) | hum, hmm, mhm | 0.55-0.6s |
| Quick single words | like, yeah, so, well | 0.25-0.35s |
| Longer single words | okay, basically | 0.4-0.55s |
| Multi-word phrases | you know, i mean, kind of, i guess | 0.5-0.55s |
Test with a few samples before processing full video.