Welcome to the atec_2018_nlp

official websit is https://dc.cloud.alipay.com/index#/topic/intro?id=3 I did some clean up, but the code is exactly what I submitted in the competition! Only one type of model used, with different input(word embedding and char embedding) you should be able to get >0.623 locally with 0.05 VP for ensemble.

• the data was for season1 only(for some security reason, season2 data can not be shared)

• to run the char model(avg score:0.5729852440408627,prd_thr:0.21000000000000002):

  python model.py -m train_model -ms "ELMOEM128_CHAR"

• to run the word model:(avg score:0.5582690659811482,prd_thr:0.18000000000000002)

  python model.py -m train_model -ms "ELMOEM128"

• to run ensemble:

  python model.py -m train_model -ms "ELMOEM128_21,ELMOEM128_CHAR_21"

• to debug:

  python model.py -m train_model -ms "ELMOEM128_CHAR" -d

• with different validation size:

  python model.py -m train_model -ms "ELMOEM128_CHAR" -vp 0.05

I do read lots papers, but only one was really used for my code(ELMO)

Deep contextualized word representations

UWB at SemEval-2016 Task 1: Semantic Textual Similarity using Lexical, Syntactic, and Semantic Information

Injecting Relational Structural Representation in Neural Networks for Question Similarit