1 件の公開記事
Compare WhisperTranscribe.ai with Sider, Transkriptor, Otter and Fireflies for a one-hour recording, source-linked briefs and timed caption handoffs.
Compare new-recording transcription, existing meeting search, summaries, exports and setup across four ChatGPT transcription integrations.
Compare 120 half-hour meeting recordings: AccurateScribe’s $19.99 subscription, Fireflies’ upload allowance, Happy Scribe costs, file limits and export options.
Compare AccurateScribe, Otter, Sonix, Notta, Descript and Rev on pricing, file limits, exports and published transcription tests.
Character error rate
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Semantic accuracy
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Names and proper nouns
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Numbers and dates
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Prices and currencies
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Addresses and identifiers
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Acronyms and abbreviations
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Technical terminology
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Punctuation
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Capitalization
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Sentence boundaries
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Paragraph boundaries
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Negation and contradictions
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Omitted speech
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Inserted speech
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Substituted words
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Silence hallucinations
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Music hallucinations
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Repeated phrases
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Long-form consistency
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Rare words and loanwords
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Confidence calibration
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
Human correction effort
Use a checked ground truth and document normalization, sample scope and denominators; distinguish measured errors from model confidence.
Proposed format: Language & accuracy tests. Publish findings only after the review is complete.
別の評価項目を選ぶか、予定トピックをご覧ください。
