BI商业智能>> disaster_tweets>> 返回
项目作者： willyiamyu

项目描述 ：
Submission for Kaggle Competition "Real or Not? NLP with Disaster Tweet" (Top 10%) using SVM's, LSTM's, and BERT
高级语言： Jupyter Notebook
项目地址: git://github.com/willyiamyu/disaster_tweets.git
创建时间： 2020-09-04T08:40:49Z
项目社区：https://github.com/willyiamyu/disaster_tweets
开源协议：
下载


Submission for Kaggle Competition “Real or Not? NLP with Disaster Tweet” (Top 10%)

Link to challenge: https://www.kaggle.com/c/nlp-getting-started/overview

Link to Public Kaggle Notebook (SVM’s): https://www.kaggle.com/willyiamyu/disaster-word-embeddings-svm

In this repository you will find 3 notebooks:

  • One using spaCy word vectors and SVM’s
  • One using BiLSTM’s
  • One using pre-trained BERT on sequence classification

SVM’s achieve f1 score of 0.81152 on the test set, BiLSTM achieve 0.80, and BERT achieves ~0.84 f1 score.