imperialite / filipino-linguistic-extractors Goto Github PK

This repository contains scripts for extracting linguistic features from Filipino texts. The scripts were created for Joseph's MSCS thesis in readability assessment of children's books. The complete list of linguistic features including the formulas and descriptions are uploaded with this repo. I advise you to check the document first before running the codes.

The scripts only contain functions for extracting a specific feature. So, you only need to create a main.py file and import the necessary script you need and call the functions. For TRAD, SYLL, and LM, I'm fairly certain you are not going to encounter any dependency issues as most scripts just rely on string manipulation. However, I you want to use LEX and MORPH, you need to setup the the following:

• JDK8 or any latest-ish version of JDK should work.
• Lastest version of Stanford POS Tagger from the CoreNLP suite. Make sure to read how to set this up on your device.
• Download the two Filipino models for the POS Tagger from Go and Nocon (2017)'s paper here and load them by reading the instruction at Stanford's FAQ website.

The scripts uploaded were customized to the needs of the previous research where the these were created. You are free to change or tinker with some of the code according to your own research. For example, in LEX and MORPH, I don't calculate features for all sentence but only for a random subset. You may change this as you like but take caution that it might take a long time to finish parsing.

You may also update some of the features if you feel like it. For example, for extracting language model features in LM, I used an old literal way of calculating perplexity by scratch derived from this repo. This can be easily done efficiently with some open-source library like NLTK or Spacy, I believe.

If you find this repository useful, please cite the following papers:

Imperial, J. M., & Ong, E. (2021). Diverse Linguistic Features for Assessing Reading Difficulty of Educational Filipino Texts. arXiv preprint arXiv:2108.00241.

Imperial, J. M., & Ong, E. (2020). Exploring Hybrid Linguistic Feature Sets To Measure Filipino Text Readability. In 2020 International Conference on Asian Language Processing (IALP) (pp. 175-180). IEEE.

Imperial, J. M., & Ong, E. (2021). Application of Lexical Features Towards Improvement of Filipino Readability Identification of Children's Literature. arXiv preprint arXiv:2101.10537.

If there is something you want to tell me about, you may contact me using the following information:

Joseph Marvin Imperial
[email protected]
www.josephimperial.com