The usage of '<oov>' is not consistent with the paper

jind11 / TextFooler

A Model for Natural Language Attack on Text Classification and Inference

MIT License

485 stars 79 forks source link

The usage of '<oov>' is not consistent with the paper #39

Open plasmashen opened 3 years ago

plasmashen commented 3 years ago

In paper, the importance score of the word is calculated by removing this word, but you use '\<oov>' to replace this word to calculate the importance score in https://github.com/jind11/TextFooler/blob/master/attack_classification.py#L216

Moreover, the '\<oov>' will be tokenized into 4 tokens which may have attention affects with other words. I'm wondering why such nonsensical '\<oov>' is used?

jind11 commented 3 years ago

hi, I have tested both methods: removing the word or replacing it with "" and the difference is not obvious. is in the vocab so I don't think it can be tokenized into 4 tokens. Let me know if you have more questions.

Youoo1 commented 2 years ago

Where is the emdding.npz file, please? Or how is it generated? 7a678cd5f2a8398b7980d8aaa9d5aec b9069123768ea397299dc7ed1419901

jind11 commented 2 years ago

The readme file has explained how to obtain the embeddings: Run the following code to pre-compute the cosine similarity scores between word pairs based on the counter-fitting word embeddings [https://drive.google.com/file/d/1bayGomljWb6HeYDMTDKXrh0HackKtSlx/view].

python comp_cos_sim_mat.py [PATH_TO_COUNTER_FITTING_WORD_EMBEDDINGS]