THL Toolbox > Audio-video > Linguistic Processing - Transcription, Annotation And Translation > Transcription & Inqscribe
Transcription Using InqScribe
Contributor(s): Bradley Aaron
InqScribe is an easy to use but powerful transcription software made by 
Inquirium. THL has adopted InqScribe (pronounced Ink Scribe) as our main transcription solution and we recommend partners use it too. However, transcription can be done using even a word processor as long as the conventions listed below are followed and time codes exactly match the A/V title that is being transcribed.
The speech of the finished title should be transcribed in the original language's script and time coded such that each clause or sentence is marked in terms of beginning and end times. Serious transcription is a difficult and time consuming process and must be done with care. Once the original speech is transcribed, the transcript then can be translated into relevant target languages. Another benefit of transcription is that one can also add rich annotations to the transcript to provide relevant details on lexical, grammatical, social and other issues pertaining to the title and its language. Intellectual processing in the form of annotations is necessary to contextualize and interpret the titles.
General transcription procedures for transcribing Tibetan language can be found in the Transcription Manual.
Using InqScribe
(Note: all instructions below are for the mac version, hotkeys may differ on Windows machines, but all commands can be found in the menu bar at the top of the screen.)
To start a new transcription project, first go to File-->New Document with Media File Then select the media file (this can be a video or audio file), the media will be loaded on the left side of the screen and a black space to create your transcription will be available on the right.
You will need to setup the fonts for the language you are transcribing. For transcripts made in Tibetan font you must use a unicode Tibetan font only, and THL requires the use of Jolmolhari-ID Unicode Font on the Mac, and Microsoft Himalaya font on Windows (Click here for more information). It is very important to ONLY use these two approved unicode fonts, as other fonts will not display correctly on all systems.
Transcript-->Transcript Settings-->Font (select Jolmohari-ID, which you can download here)-->Size (change to 21, or bigger if needed)
You also need to make sure that the frame rate settings are set to milliseconds so that mediabase will have the best possible match for the transcription. To do this, go to InqSCribe--> Preferences--> New Document --> Frame Rate: and change to "1000 fps (Milliseconds)" and click "ok".
The goal is to create a transcript of what is said that, at a minimum, registers the time code from the original media file and includes the speakers' names. Timecodes should be entered into the transcript according to this guide.
The interface for Inqscribe is customizable, but the default keyboard shortcut to start/pause the media play is the "Tab" key, you can also go back 8 seconds in the media file by hitting "control+Tab". The other useful hotkey is to insert a time code, this can be done by hitting "command+;"
For songs or poetry where only one person is singing, you do not need to mark a speaker and should JUST enter the timecode by using the hotkey. It will look like this [00:01:00.000] འདི་བོག་ཡིག་རེད་།
For transcripts where there are people speaking, you also need to mark each new speaker (except in cases where only one person is speaking. You only need to mark speaker names when more than one person speaks in the audio or video). This is done by putting the name of the speaker, followed by a colon (ex. Speaker 1: ) before each sentence of every speaker (this is facilitated by using inqScribe 'snippets'. See below for details). If you are entering bilingual speaker names (for example a Tibetan and English rendering of the name) you use this format "བོད་ཀྱི་མིང་། / English Name:" You MUST use the "/" between the Tibetan and the second language name. The second language can be English, Chinese or another language.
Your transcript should basically look like this for only one language speaker names (time codes are in HH:MM:SS.SSS format):
[00:00:00.000] Name of speaker 1:
This is a placeholder for the transcribed text that you will be putting into the transcription text box. you will put everything the speaker says here in the language used in the video.
[00:01:00.000]
[00:01:11.010] Name of speaker 2:
This is where you will put what speaker 2 is saying, and you will do this for every speaker according to the guide.
[00:10:08.010]
It should be automatic if you are using snippets, but if you are entering speaker names manually, make sure you include a space between the time code closing bracket and the first letter of a speaker name or transcript.
You do not need to enter the closing 'out' timecodes if you are planning on using the file on UVA's Audio-Video App., when you export as XML, you can have InqScribe add the out points automatically.
To make entering speaker names easy, you can set up what in InqScribe is called a "snippet" to make the entry of speakers and time codes much quicker by assigning a set string of text to a hotkey. To setup snippets for all speakers in a transcript go to Edit--> Edit Snippets and then click "Add". You will then be able to create a new snippet which you should name "Speaker 1" and set the snippet text by clicking in the Snippet: code box and entering this text: "{$TIME} Speaker 1:". Replace "Speaker 1" with the name of the first speaker whose speech you are transcribing. This should be in the language of transcription. (ex. {$TIME} བཀྲ་ཤིས:) and you will have to change the snippet text for each new transcript where the speakers' names are different. Once you've created the snippets for all speakers, you can define a hotkey to use to easily enter the snippet into your transcript. To do this, click "Define Trigger" in the Edit Snippet menu and then hit the hotkey or combination of keys you want to use to automatically enter the snippet into your transcript. On Windows machines, F2 is a good key to assign to the speaker 1 snippet, and then F3, F4, etc. for each additional speaker. You only need to set your trigger once and it will save the trigger for each new transcript, but you still have to edit the snippets to change the names of the speakers for each transcript.
If you have translations in another language different from the language of the transcriptions, put the translations after each native language transcription on a new line and within the time codes. Translations should be done AFTER the initial transcription. You can have translations in multiple languages such as English and Chinese in the same transcript, just make sure that you do not mix character sets in the translation. This means that if you write in English, do not put Tibetan characters mixed in with the English translation, and if you write in Chinese characters, don't put Tibetan or English characters mixed in with the Chinese (ex. Don't mix characters in translations like this: This is Tibetan, which is བོད་ཡིག་ in Tibetan characters. OR 这是第2件项目.)
For example:
[00:01:11.010] Speaker 1:
འདི་བོག་ཡིག་རེད།
这是藏文。
This is Tibetan.
[00:01:13.020]
Annotating Transcripts Inline
It is possible to include annotations in your transcripts that help to contextualize the transcribed materials.
To include annotations, use this format:
[00:01:11.010] Speaker 1:
འདི་བོག་ཡིག་རེད།
This is Tibetan.
<འདི་བོག་ཡིག་རེད། Contextual annotation.>
[00:01:13.020]
You should save the project as the native inqscribe format with the same name as the video file that you are transcribing (this convention is VERY important, and MUST be followed) and it should be in a folder near the folders holding the media files to keep from losing track of the transcription files.
The final transcript/translation file should also be exported as an XML file, which can then be uploaded to Mediabase and matched with the A/V file. Here are instructions on how to export the transcript in the proper format for the online A-V database.
Dealing with Rapid-fire Changes in Speakers
Sometimes, when there is more than one speaker in the media you will encounter a rapid change between speakers or even overlap between speakers. You should continue with the convention of marking each speaker in sequence, even if there is overlap. So, a transcript with two speakers talking at the same time would look like this (note that the time codes are the same, but the text and speaker label is different:
[00:01:11.010] Speaker 1:
འདི་བོག་ཡིག་རེད་བས།
[00:01:13.020]
[00:01:11.010] Speaker 2:
རེད།
[00:01:13.020]
Working with Transcripts outside of InqScribe
Transcriptions can be changed outside of InqScribe in text editors, saved as plain text (*.txt) and then imported as 'plain text' into InqScribe as long as all conventions listed above are followed. This also means that it is possible to export an InqScribe file as a plain text file (complete with time codes and any transcriptions) and then make changes in a text editor, such as the addition of translations to existing original language transcripts, and then import the text file back into InqScribe and save as an InqScribe file with all time codes and transcriptions intact, which can then be exported as XML for Mediabase.
Working with Transcripts from Mediabase Offline in InqScribe
You can also export transcripts already in Mediabase as text files so you can work on them offline in InqScribe, make changes, and then re-upload to Mediabase using the procedure above. For now, to do this, you must enter a custom URL and change the indicated "node=" portion to the mediabase ID for the transcript you want to export.
For example, to export the transcript for the Mediabase title "A Lucky Dream: Three's Company #01" you would put the following URL into your browser, which would initiate a download of a text file of the transcript in a format that can be read by InqScribe:
http://mediabase.drupal.shanti.virginia.edu/inqscribe-export?node=185
To download the transcript of other Mediabase titles, first find the title for which you want to export the transcript, click "Technical metadata" and copy the "Shanti AV ID" number and paste it into the URL above after "node=".