Viewing a transcript

A transcript is a collection of annotations on numerous layers corresponding to a single audio file, plus transcript attributes like when the audio file was originally recorded. In APLS, each transcript has a single main participant and corresponds to a section of that participant’s sociolinguistic interview. Each transcript has its own page in APLS.

On this page, we use the term transcript for the underlying data in the APLS corpus, and transcript page for the webpage you can use to view transcript data.

As we’ll cover below, there are a few ways to access individual transcript pages. The easiest is the transcripts page at https://apls.pitt.edu/labbcat/transcripts. Go to that page and select a transcript to view.

On this page
  1. What you can do on the transcript page
  2. Transcript page layout
  3. Media player and navigation arrows
  4. Transcript header
    1. export formatted ▼
    2. [transcript attributes] panel
    3. Participants list
    4. Layer selector
  5. Transcript body
    1. Turns, utterances, noises, and overlaps
  6. Accessing transcript pages

What you can do on the transcript page

Transcript page layout

Individual transcript pages have two sections, plus some elements that remain in the same position on the page as you scroll. This is highlighted in the following screengrab:

  • Sections:
    • The transcript header, highlighted in red. This includes information about the transcript, a menu for exporting the transcript in various formats, and settings for choosing which layers are displayed in the transcript body.
    • The transcript body, highlighted in blue. This displays the actual content of the transcript. It continues all the way down the page.
  • Fixed elements:

Media player and navigation arrows

There are two features of the transcript page that are fixed elements, meaning that they remain in the same place on the page regardless of where you scroll. These elements are the in-browser media player and the transcript navigation arrows (see previous screengrab).

The transcript navigation arrows allow you to move from one transcript to the either the previous transcript or the next transcript in the interview. These arrows are hidden if you are looking at the first or last transcript in the interview.

On the transcript page for CB01interview3.eaf, click the “previous” button. This will load the transcript page for CB01interview2.eaf. Then, click the “next” button to go back to CB01interview3.eaf.

The URL for a transcript page looks a little different when you use the “previous” or “next” buttons. For example, the URL for CB01interview3.eaf ends with ?ag_id=354 (rather than ?transcript=CB01interview3.eaf) when it’s reached from the “previous” or “next” button. The ag is short for annotation graph, the data model that LaBB-CAT corpora are based on. For more information, see Bird and Liberman (1999) and Fromont (2017).

The in-browser media player allows you to listen to audio from the transcript, control the volume, and (via the ⋮ menu) control the playback speed. While the audio is playing, the transcript body will highlight the utterance (line) that is currently playing. You can also listen to audio through the word menu.

If your media player is displaying a small Praat logo (as shown below), this indicates that you do not have the Praat integration set up. You can learn more about enabling the Praat integration here.

Transcript header

The transcript header provides information about the transcript, options for exporting a formatted transcript, and check boxes that control what data is displayed in the transcript body.

  • At the top of the transcript header is the name of the transcript file (in this case, CB01interview3.eaf).
  • Most important is the layer selector, highlighted in red. This controls which layers are displayed in the transcript.
  • export formatted ▼, highlighted in blue: a drop-down menu that allows you to download the current transcript in multiple formats.
  • [transcript attributes], highlighted in green: a collapsible panel that displays transcript attributes.
  • A list of participants, highlighted in purple, with links to attributes pages.

export formatted ▼

The export formatted ▼ drop-down menu allows you to download the current transcript in any of the listed file types:

On the CB01interview3.eaf transcript page (https://apls.pitt.edu/labbcat/transcript?transcript=CB01interview3.eaf):

  1. Make sure only the word layer is selected.
  2. Click export formatted ▼.
  3. Click Praat TextGrid.
  4. Open the TextGrid in Praat.

At a minimum, the exported file will include the utterance and word layers. It will also include any layers you have selected to show in the transcript body. Some transcript formats also show additional layers from the original uploaded transcription: noise, comment, and pronounce.

If you need greater control over which layers to export, use the Export Formatted menu on the transcripts page.

The file format affects not only which layers are included, but how they are displayed. For example, in Praat TextGrids and Elan EAF Transcripts, there is one tier for each speaker’s utterance annotations (labeled, for example, Barbara Johnstone tier and one CB01 tier) and one tier for each speaker’s word annotations (e.g., word - Barbara Johnstone and word - CB01); these annotations are all time-aligned. In Plain Text Documents and PDF Documents, utterances are separated by line breaks and words are simply listed within each utterance; these annotations are not time-aligned and there’s no way to tell whether speech is overlapping. As a result, the Plain Text Documents and PDF Document formats are better for a readable or skimmable overview of who said what, and not very useful as structured linguistic data.

  1. Open the TextGrid from the previous try-it.

You should see four tiers, two per speaker: one for each speaker’s utterance annotations (labeled Barbara Johnstone and CB01) and one for each speaker’s word annotations (labeled word - Barbara Johnstone and word - CB01).

Now let’s zoom into the first Barbara Johnstone utterance (okay just a few questions about you as an individual):

  1. Click that interval in the Barbara Johnstone tier.
  2. Either click sel in the bottom-left of the TextGrid window, or hit <ctrl>/<cmd>+N.

You should see that each of the words in the utterance has its own interval in word - Barbara Johnstone:

Finally, let’s see how overlaps look in an exported TextGrid. According to the transcript page, there’s an overlap when Barbara Johnstone says what you'd say you're and CB01 says okay, right after Barbara’s first utterance

  1. Zoom into the next Barbara Johnstone interval in the TextGrid.

You should see a CB01 interval that lines up with the Barbara Johnstone interval, plus words on both speakers’ word - tiers.

As previously mentioned, the exported file will include any layers you have selected to show in the transcript body. Unlike in the transcript body, phonological layers (like segment) are always displayed in the DISC phonemic alphabet.

On the CB01interview3.eaf transcript page:

  1. Select the segment layer.
  2. Click export formatted ▼.
  3. Click Praat TextGrid.
  4. Open the TextGrid in Praat.

Compared to the previous try-it, you should see two new tiers: a segment - tier for each speaker. If you zoom in on an individual word, you should see that each of the segments has its own interval.

Finally, if a layer allows vertical peers (with a icon in the layer selector), the exported file may show multiple tiers per speaker:

On the CB01interview3.eaf transcript page:

  1. Unselect segment and select dictionary_phonemes.
  2. Export to Praat TextGrid and open in Praat.

You should see three dictionary_phonemes - tiers per speaker. This is because some words have more than one possible phonemic representation, such as /æz/ or /əz/ for as.

[transcript attributes] panel

Underneath the export formatted menu is the collapsible [transcript attributes] panel This panel shows’ transcript attributes display titles in the left-hand column (e.g., Neighborhood) and values in the right-hand column (e.g., Cranberry Township).

If you hover over a display title, a tooltip will pop up with the attribute’s export name and short description.

Blanks in the right-hand column mean that this transcript doesn’t have any value for that attribute. For example, the previous screengrabs show a blank for Transcription AI tool(s) because the transcription_ai_tools attribute is empty for CB01interview3.eaf; no AI tools were used to transcribe that interview.

To export transcript attributes to a CSV file, use the Export Attributes menu on the transcripts page.

Participants list

Below attributes are the participants in the transcript.

Participant codes for main participants are displayed in bold. Clicking a participant code will take you to that participant’s attributes page.

Layer selector

The projects and layers sections control what information is displayed in the transcript body.

The word layer cannot be turned off (which is why its checkbox is grayed-out).

To make more layer options appear in the layers section, the project that corresponds to that layer must be turned on in the projects section.

You can hover over the names of the projects to view a brief description of the layers associated with that project. You can also read more about projects on the Layer typology documentation page.

As shown in the screenshot above, toggling the timing project causes the layers section to display layers that are included in the timing project category. You can learn more about layers in the Layers and attributes documentation pages.

The layer options are ordered by scope from largest to smallest, then by project in the order shown in the projects column, and then by alphabetic order.

Layers with phonological information have an additional toggle to switch between IPA transcriptions and Raw (DISC) transcriptions. You can read more about DISC and why APLS uses it on the Phonemic transcription documentation page.

Selecting one or more of the options in layers section will cause that layer to be displayed in the transcript body.

Transcript body

To become acquainted with the layout of the transcript body, we will start off with only the default word layer enabled, as shown in the screenshot below. Additionally, while the other layers affect the information that is displayed in the transcript, all the interactable layers in the transcript body operate in the same way as the word layer described in this section.

Turns, utterances, noises, and overlaps

Every line in the transcript corresponds to a conversational turn or continuous utterance in the audio recording. The speakers who have transcribed audio for a particular turn are shown on the left-hand side within the dotted lines. In the example screenshot above, the interviewer Barbara Johnstone is the only person speaking in these lines, so only the name Barbara Johnstone is displayed.

To the right of the dotted lines is the transcribed speech for that speaker for that utterance. Clicking on any of the blue words in the transcript will bring up the word menu that allows you to open the utterance in Praat, extract and download the audio of the utterance, or play the audio of the utterance using the in-browser APLS media player.

For instructions on setting up the Praat integration for your web browser, visit the Viewing transcript segments in Praat documentation page.

When an utterance has overlapping speech from multiple speakers, the left-hand side of the transcript will show all the speakers who have transcribed speech for that portion of audio. Participant speech is bolded in APLS to make it easier to distinguish from the interviewer’s speech.

If non-speech sounds have been transcribed, those annotations will appear above the word layer. All non-speech sounds from all speakers are combined in the top-most section of the word layer, so it is possible for a speaker to appear on the left-hand side of the transcript without having any words in the word layer. An example of this is shown below.

It is possible to use CTRL+F to search a transcript page for singular words but not for multi-word strings due to the underlying HTML of the transcript pages (e.g., CTRL+F will work to find the word ‘filing’ but it won’t find the phrase ‘filing cabinet’).

Accessing transcript pages

There are multiple ways to reach a transcript page in APLS:

In addition, there are two ways to access transcripts from the results of a search (see Searching the corpus):

  • Through the search results page by clicking on hyperlinked text. This will load the transcript page and scroll directly to the search result you clicked.
  • If you export search results to CSV (see Exporting data), the URL column has links that load the transcript page and scroll directly to the search result corresponding to each row.