junwei-pan / fasttext_benchmark

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

fasttext_benchmark

Test the fasttext model on text classification over several datasets provided in Xiang Zhang's paper Character-level Convolutional Networks for Text Classification(Download Datasets from Google Drive), the code is based on the fasttext example on keras: keras fasttext example.

Experiment results

  • ag_news
embedding_dim 1 epoch 2 epochs 3 epochs 4 epochs 5 epochs
20 84.89% 86.70% 87.86% 88.54% 89.50%
50 86.29% 88.01% 89.34% 89.92% 90.32%
  • dbpedia
embedding_dim 1 epoch 2 epochs 3 epochs 4 epochs 5 epochs
20 94.06% 96.36% 97.05% 97.40% 97.56%
50 96.00% 97.23% 97.61% 97.76% 97.86%

Reference

  1. Bags of Tricks for Efficient Text Classification
  2. Character-level Convolutional Networks for Text Classification
  3. keras fasttext example

About


Languages

Language:Python 100.0%