Amazon Transcribe is misplacing punctuation (period) and incorrectly moving words to next/previous sentences.

0

While attempting to review the Transcribed Closed Caption, we found a frequent and consistent error where the punctuation is applied in the wrong place. We reviewed the video(s) carefully and found that the audio is very clear and concise where one sentence end and another begin. Despite that, we see punctuation errors.

We are not exactly sure, how can we solve this problem. Is there an existing mechanism to deal with such an issue?

Examples of missed Punctuation: Wrong 1: … who are unloading the shipment use 40 inches by 48 inches... Correct 1: … who are unloading the shipment. Use 40 inches by 48 inches…

Wrong 2: … or FMVS test units that have moving…

Correct 2: … or FMVS test. Units that have moving…

Wrong 3: … and when stored liquid products must adhere…

Correct 3: … and when stored. Liquid products must adhere…

Wrong 4: … the edge of the pallet. More than one inch

Correct 4: … the edge of the pallet more than one inch…

질문됨 일 년 전322회 조회
1개 답변
0

Amazon Transcribe is a machine learning-based service and may occasionally make errors, especially when it comes to punctuation and capitalization. However, there are some ways to mitigate these errors:

  1. Review the transcribed text carefully and correct any errors manually. This can be time-consuming, but it will help ensure the accuracy of the text.
  2. Use a post-processing tool to correct punctuation errors automatically. For example, you could use a regular expression to search for common punctuation errors and correct them automatically.
  3. Consider using a punctuation-aware language model or tool to improve the accuracy of the transcribed text. There are many language models available that are specifically designed to handle punctuation and capitalization errors.
  4. Provide feedback to Amazon Transcribe. If you notice consistent errors or issues with the service, you can provide feedback to Amazon so that they can improve the accuracy of the service in the future.

Keep in mind that even with these measures, there may still be occasional errors in the transcribed text. It's important to review the text carefully and make any necessary corrections to ensure the accuracy of the final output.

hash
답변함 일 년 전

로그인하지 않았습니다. 로그인해야 답변을 게시할 수 있습니다.

좋은 답변은 질문에 명확하게 답하고 건설적인 피드백을 제공하며 질문자의 전문적인 성장을 장려합니다.

질문 답변하기에 대한 가이드라인

관련 콘텐츠