asc-analyzer 0.1.1

asc-analyzer

ASC-analyzer

The ASC analyzer extracts Argument Structure Constructions (ASCs) from raw English texts and computes indices related to ASC usage.

Installation

To ensure stability, I recommend installing dependencies in the following order:

1. Install spaCy:

   pip install spacy
   

2. Install spaCy-transformers:

   pip install spacy-transformers
   

3. Download the transformer-based spaCy model:

   python -m spacy download en_core_web_trf
   

4. Install the ASC analyzer package:

   pip install asc-analyzer
   

Quickstart

Prepare a directory with .txt files (e.g., data/text/). Each file should contain plain English text.

Then run:

asc_analyzer \
  --input-dir data/text \
  --source cow \
  --print-asc \
  --save-asc-output

This command will:

• Assign ASC tags to each sentence
• Print the ASC-tagged results directly to the terminal (--print-asc)
• Save token-level ASC tagging results as *_ASCinfo.txt files (--save-asc-output)
• Compute ASC usage statistics (e.g., diversity, proportion, frequency, and verb–ASC association strength) and save them in a CSV summary file
• The --source option determines which reference corpus is used for computing frequency and association measures:
  • cow: uses the COW corpus (web-based, written English)
  • subt: uses the SUBTLEX corpus (subtitle-based, spoken English)
  • Choose the source based on the register that best matches your input data.

Options

Option	Description
--input-dir	Directory containing .txt files to process (default: asc_analyzer/data/test)
--output-csv	Path to save the resulting CSV (default: Written_COW.csv or Spoken_SubT.csv)
--source	Reference dataset: cow (written, default) or subt (spoken)
--indices	Comma-separated list of index names to include in the CSV (default: all standard indices)
--save-asc-output	Save ASC-tagged outputs as *_ASCinfo.txt in the input directory
--print-asc	Print ASC-tagged results to the terminal

Output for --print-asc

When using the --print-asc option, the output for each sentence shows aligned token information and its ASC label (None if no ASC applies):

# sent_id = 1
1	The	the	
2	idea	idea	
3	is	be	ATTR
4	trust	trust	

You can save this output to txt files by including --save-asc-output.

Citation

• If you use the ASC tagger (i.e., --print-asc, --save-asc-output), please cite:

  • Sung, H., & Kyle, K. (2024). Leveraging Pre-trained Language Models for Linguistic Analysis: A Case of Argument Structure Constructions. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP).

• The ASC analyzer is currently in beta testing and will be updated soon.

License

This project is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.