Parsing_Speech

Dependency Annotation and Parsing for Spontaneous Speech

The current directory contains data, code and models for Dependency Parsing Evaluation for Low-resource Spontaneous Speech

1. Convert original CHILDES xml file to .conllu file
   1. python3 code/ori_xml2conll.py --input Path_To_Corpus --output Path_To_Output --section Section (e.g. English-NA or English-UK)
   2. eve.py tailored specifically for the Eve corpus from the Brown corpus
2. Semi-automatic conversion from CHILDES annotation to UD annotation
   1. python3 code/converter.py --input Input_Path --output Output_Path
   2. in data/Eve/eve_annotated
3. Manual annotation
   1. in data/Eve/eve_annotated
4. Significance testing of parsing results
   1. python3 code/bootrap.py --gold Gold_Annotation_File --pred Predicted_File --n Number_Of_Iterations (e.g. 10000) --c Sample_Size (e.g. number of utterances in the file)
5. Descriptive statistics of child information from CHILDES
   1. python3 code/descriptive_statistics.py --input Input_Path --output .csv_file
   2. English: in results/en_descriptive.csv
   3. Chinese: in results/zh_descriptive.csv
6. Models