Speechnotes is a good program for newcomers, but Nuance Dragon Professional is one for the pros. It has one of the highest price points of any ASR solution but it also offers accuracy worth paying for. Few other solutions can match this application in terms of accuracy, and it has flexible price points to suit modest budgets.

Thank you for the information you have provided. Nishat Vasker I will check them out.

What languages will you be studying, Ilknur? The language and the subject matter affect the quality of ASR. Dragon, for example, is not offered in most world languages, but as Nishat says, it is excellent in those languages it does support. The PC version only. The Mac version sadly is riddled with errors unless they have fixed this.

Dear Sue Reager , I am a master's student in the department of English Language Teaching. Therefore, I am studying English. However, some speakers' English does not sound as well as native speakers in some parts of the videos or voice-recordings since they are learners of English in various levels including A1, A2, B1 and B2. Almost half of the recordings, mostly the teacher's voice is recorded, which sounds much better than the students; easier to be understood by ASR systems I guess. I thank you so much for your response. Could you give me a little bit more details about Dragon, please? Is it free to use or should I subscribe by paying some fee? If it's not free, I hope it is not too expensive. (By the way, I use a laptop using Windows 10 pro software, not a Macbook.)

I have done extensive testing of ASR engines with accented English and other languages (so far, about 15 engines). The top for accented English are Dragon, Speechmatics, Microsoft and rev.ai (for subtitling).

DRAGON: Dragon has a way to select the accent of the speaker, but is a bit complex. Dragon assumes there is one user of their software. So, is uncomfortable to switch accents. Dragon is not free, but there is an educational price. See the website. Dragon has the advantage that you can teach it words, so if some words that are constantly in the audio that you need to transcribe are not in the Dragon corpus, you can add. Dragon has an interface and ability to import mp3 (in the past, it did not import video, you will need to review the updates to the package to see if has changed).

MICROSOFT: Excellent with accents. Has improved beyond recognition compared to last year. I believe that you can set up a student account with Microsoft to use their API. This means, creating your own mini-interface to handle your data. Microsoft has an online version, but the online free Microsoft often makes errors whereas the API version does not make those errors. So, if you know how to make your own interface, Microsoft will probably be exactly right. And you may be able to use Microsoft for free if you are clear about the educational goal. I cannot speak for the company, of course. Perhaps you can offer extensive content analysis to help them improve their own product.

SPEECHMATICS: Excellent with accents. Highly - highly - recommend. You may well need to make your own interface for this, as well. I don't know if there is an online version - we use APIs. Contact them to ask about cost. Will not be free, but they may well have an educational offer.

REV.AI: Very good, but not free. It auto-creates subtitle files with timecodes and this may be interesting to you.

In summary, if you are doing serious study that will help any ASR company improve, they may be open. Identifying errors with a professional eye is important to any good ASR company.

Thank you so much for your very detailed answer Sue Reager . I have never thought about the ASR systems of the companies demanding high fees before reading your recommendation, but now I have seriously begun to think about giving them a chance by demanding free access to their software via e-mail and offering to help them develop their company regarding the ASR.

I have used Otter.ai for the transcription of data of my thesis study. Otter does not offer transcriptions for languages other than English, which in my case it was a mix of Eng-Tr, but, as for my experience for English transcription, it achieved a %80-95 accuracy , the lowest percentage was mostly due to audio problems or speaker's inaccurate pronounciation of words. After the software completes the transcription process, you definitely need to check yourself for the accuracy, however, it saves a lot of your time considering you don't need to transcribe manually. Besides, it is not so expensive. Otter has budget options for student accounts.

Thank you for your detailed suggestion Esra Çil Hocam. I will definitely check it out!

I'd go for Google and Microsoft since they have some of the largest data bases of speakers of English from various regions therefore they can easily process the phonemic differences such as t and th and w and v for Turkish speakers of English.

Thank you Umut Salihoğlu Hocam! I will give them a try.