Look up a word you can't remember by typing a plain-language description of what it means in English or Chinese.
Run a local version of the reverse dictionary offline by deploying the Django backend with the pre-trained PyTorch model.
Integrate the reverse-dictionary model into your own NLP research pipeline or writing tool.
Requires downloading large pre-trained model files from an external link and pinning specific PyTorch and dependency versions.
WantWords is an open-source reverse dictionary. A regular dictionary takes a word and gives you its definition. A reverse dictionary works in the opposite direction: you describe a concept or idea in plain language, and it returns a list of words that match what you described. This is useful when you know what you want to say but cannot think of the right word, a situation sometimes called the tip-of-the-tongue phenomenon. The project is available as a live website and also as a WeChat mini-program. It supports both English and Chinese queries. The underlying model is a multi-channel neural network trained to map descriptions to semantically matching words. The model was published as a research paper at an academic AI conference and the code for that model lives in a separate linked repository. To run the system yourself, you download pre-trained model files and data from a provided link and place them in the correct directory. The backend is a Django web application, and the system depends on PyTorch and several other Python scientific computing packages. The README lists specific version requirements for all major dependencies. This repository comes from the Natural Language Processing Lab at Tsinghua University. The README asks users who use the code or data in their own work to cite the two associated research papers. The project is described as open-source, though the README does not specify a license.
← thunlp on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.