Nana2929 / 2021Fall_Computational_Semantics_Project

2021 Fall: Computational Semantics Term Project

Chinese Wiki-Based Word Sense Disambiguation

1. Topic

   Chinese Wiki-Based Word Sense Disambiguation

2. Data

   • Crawled from Wikipedia, cleaned and preprocessed by myself
   • 164 words with # of fine-grained senses >=2

3. Phrase list: https://gist.github.com/indiejoseph/eae09c673460aa0b56db

4. The Experimented WSD Algorithms

   (1) tfidf (inspired by https://www.itread01.com/hkyecfy.html)

   (2) Lesk++ (inspired by Simple Embedding-Based Word Sense Disambiguation)

5. Result

   (1) tfidf

   • Train: Test = 1:1, Total TestSize = 28,517 // No randomization
   • Total Accuracy: 46.6% // Coarse-grained Accuracy

   (2) Lesk++

   • Train: Test = 7:3, Total TestSize = 8,576 // with randomization
   • Total Accuracy: 59.34 % // Coarse-grained Accuracy

6. Future Expectations

   (1) cleaning and reviewing data manually

   (2) extending dataset size

   (3) application on WSD fine-grained task: named entity (exploiting features of Wikipedia)

Check doc for more details.