brightmart / nlu_sim Goto Github PK

pretrained word embedding

I got this repo working / it’s a few years old. It has the power to take a sentence / and switch out vocab based off of respective parts of speech.
https://github.com/johndpope/vocab-mashup

It seems like this functionality could be useful in this repo, in that similar sentences could be generated. Or perhaps this is already possible?

word2id 里面PAD_ID = 0，UNK_ID=1，所以embdding的赋值开始index应该是2才对，目前是只针PAD_ID 设置了zeros，而UNK_ID没有设置，开始index是从1开始。

thanks for sharing this good work!
how much will these features improve your f1 score?

1)n-gram similiarity(blue score for n-gram=1,2,3...);

2) get length of questions, difference of length

3) how many words are same, how many words are unique

4) question 1,2 start with how/why/when(wei shen me,zenme，ruhe，weihe）

5）edit distance

6) cos similiarity using bag of words for sentence representation(combine tfidf with word embedding from word2vec,fasttext)

7) manhattan_distance,canberra_distance,minkowski_distance,euclidean_distance

https://dc.cloud.alipay.com/index#/topic/intro?id=3
排名大概如何？

找不到pretrained word embedding in data\asttext_fin_model_50.vec该文件呀