apls-v2.0.0

The checker now accepts Praat TextGrid files as input, in line with the evolving acceptance of Praat transcriptions for APLS. As such, the app has been renamed to "Transcription Checker", reflecting its broader applicability. Under the hood, this has meant changing the code to make it more flexible and consistent (continuing the changes in apls-v.1.4.0), such as defining a new function as.trs_transcription() as an S3 generic that accepts both Elan-based and Praat-based data structures. I found that it's orders of magnitude faster to call Praat directly (with a custom Praat script) than to use the readtextgrid R package, so the app now ships with Praat (for Windows and Linux) under its own license (GNU GPL).

Full Changelog: apls-v1.4.0...apls-v2.0.0

apls-v1.4.0

Changes relative to apls-v1.3.0

The code has been overhauled with a better and more consistent API. Transcriptions are represented as nested lists of tier dataframes from the get-go, with R attributes representing XML attributes and other tier metadata. This makes it much easier to run the code interactively, and will pave the way for a format-independent implementation (i.e., the ability to upload TextGrids as well as Elan files).

In the process, I discovered that the exported files actually didn't have any overlaps fixed. This version corrects that fundamental bug.

apls-v1.3.0

Changes relative to apls-v1.2.0. (The code changes would only merit patch-versions, but I'm updating the minor-version because of the relicensing.)

• App
  • Step 1
    • Check for AUTHOR file attribute rather than ANNOTATOR tier attribute (the latter is still permitted, but its absence no longer throws an error). This allows for better alignment with LaBB-CAT upload
    • Add prohibited tiers to account for Batchalign or AI segmentation
  • Step 3
    • Add new options for dealing with zero-width annotations. Elan sometimes creates zero-width annotations that aren't intended by the transcriber (often duplicates of other annotations); since these zero-width annotations aren't visible to the transcriber, they should usually be dropped
• Repo
  • Change license from GPL to CC BY-NC-SA 4.0
• Testing
  • Add/update testfiles for new/changed cases

apls-v1.1.0

• App
  • Step 1
    • Add functionality to prohibit tiers by name
    • Bugfix to handle situation where no tiers have a participant attribute
  • Step 2
    • Much more descriptive text if Step 2 fails
    • Option to use local dictionary (inclusively or exclusively) for testing
    • Send dictionary info to console for debugging
    • Bugfixes: issue with possessive clitics, false negatives caused by hyphens, and { in pronounce codes
    • Enforce punctuation errors like "hand."
  • Step 3
    • Bugfix to bypass overlap-checking if just one speaker and no redactions
  • Appearance: Add djvill/APLS's doc.css & favicon
• Repo
  • Reorganize to remove Southland apps
  • Move dictionary to djvill/APLS
• Testing
  • Better test input files
  • Test output files no longer store volatile info (e.g., file size)
• Misc
  • Add deployment shell script & enforce FALSE for testing-related switches

apls-v1.0.2

• Add Unisyn & APLS dictionary
  • Better utilities for converting csv version of APLS dictionary (which is meant to go into LaBB-CAT)
• Fix overlap-checking bug in zero-width checking"
• Refactor initial file-reading (to be better in sync with interactive use)

apls-v1.0.1

• Fix bug w/ missing TIER_ID and eaflist_to_df
• Add GitHub link

apls-v1.0.0

Update README; Update completion wording on app