Machine Translation Testing via Syntactic Tree Pruning

File Structure

./:STP Source Code

./data/original_sentences: original sentences in the form of txt and excel

./data/evaluated_results: suspected translate errors by each tool after human evaluation

./related-documentation: relationship map used when pruning

./test: test root

Python Version

• python 3.6

Install Requested Python Packages

• pip install -r requirements.txt

Install models

• Stanford Corenlp

Testing step

• Download Stanford Corenlp service(Windows/Linux/MacOS)

  • https://stanfordnlp.github.io/CoreNLP/index.html

• Detect translation errors:

  • Set dataset, api_key in main.py
  • run main.py

• Data process

  • After detecting translation errors
  • run all_output.py

• RQ Codes：

  • edit main function,and run result/all_output
  • set overlap in main function: overlap()
  • ErrorType in main function: STPType()
  • IssueCount in main function:Pairs()
  • ErrorCount in main function:Error()