Iris (Renfen) Hu  Associate Professor Computational Linguistics Department of Digital Humanities School of International Chinese Language Education Beijing Normal University Email: irishu@mail.bnu.edu.cn Google Scholar    Github    中文页面

Bio

Teaching

Selected Publications

Resources and Applications

• Chinese Word Vectors    over 11,000 stars on Github
  
  • This project provides 100+ Chinese Word Vectors (embeddings) trained with different representations (dense and sparse), context features (word, ngram, character, and more), and corpora.
  • Moreover, we provide a Chinese analogical reasoning dataset CA8 and an evaluation toolkit for users to evaluate the quality of their word vectors. Detailed discussions on the properties of word embeddings can be found in our papers:
    Analogical Reasoning on Chinese Morphological and Semantic Relations, ACL 2018.
    Revisiting Correlations between Intrinsic and Extrinsic Evaluations of Word Embeddings, CCL 2018 Best Paper.
    
  
  
• Empower AI to understand ancient Chinese
• Understanding ancient Chinese texts is a very difficult task even for experts in this area, as it involves complex linguistic phenomena and relies heavily on historical and cultural knowledge.
• We developed a multi-task joint learning approach to build a punctuation and named entity recognition system for ancient Chinese texts, which won first place in the CCL 2020 Ancient Chinese Named Entity Recognition Competition. This system has been applied in various ancient text projects by research institutions and publishing houses.
• Recently, we constructed "AI Taiyan," an ancient Chinese large language model that demonstrates strong capabilities in interpreting classical literature, supporting various challenging tasks such as word sense explanation, classical-modern Chinese translation, and allusion recognition. This model shows great potential for applications in assisting ancient text collation, dictionary compilation, and digital humanities research.
• Furthermore, we have open-sourced a series of ancient Chinese language resources to facilitate related research, such as the ancient Chinese word sense annotation corpus and the Tongjiazi dataset.






• CCA: Chinese Collocation Assistant
  
  • CCA is a free online collocation search application to assist Chinese vocabulary learning and teaching. It has served learners and educators from dozens of countries across six continents.
  • We have open-sourced the Chinese Collocation Knowledgebase used in CCA, which was automatically extracted from the Chinese Textbook Corpus (2.4M words) and Chinese Wikipedia (138M words).
  • To better facilitate collocation research, we have developed the Chinese Collocation Analyzer, a tool that supports automatic extraction of syntactic collocations in Modern Chinese and calculation of syntactic complexity measures.
  
  
  
  
  
  
• Intelligent Language Testing for L2 Learners
  
• We proposed an innovative testing approach that integrates automatic item generation (AIG) and computerized adaptive testing (CAT) for assessing Chinese as a second language (CSL) proficiency. Our system enables a rapid, adaptive Chinese proficiency test that takes approximately ten minutes.
• Additionally, we developed L2C-Rater, an automated essay scoring tool for CSL learners. The source code for L2C-Rater is available on GitHub under an open-source license.





Diachronic Sense Modeling of English Words

• We propose a sense representation and tracking framework that aims to answer not only what and when, but also how word meanings change over time.
• We have open-sourced diachronic sense evolution data and visualization tool for 3,220 polysemous English words spanning 200 years (1810-2009).
• Furthermore, we model word change from an ecological viewpoint and identify two interesting sense behaviors in the process of language evolution: sense competition and sense cooperation.