Gabriele Pergola

Loss linear combination

2

Are the coefficients used while summing up the losses fixed? How did you choose them? In the original paper, they would be _alpha_t_, where _t_ is the particular task. Thank...

Performance on the paper's dataset

6

The performance reported in the Readme has not been computed on the same dataset used in the original paper ([Hierarchical Attention Networks for Document Classification](https://www.cs.cmu.edu/~diyiy/docs/naacl16.pdf), Tang et al. 2016). It...

Dimensionalities of word minibatch and Embedding layer don't match

1

I wonder whether there is an error due to what Pytorch is expecting as input to the nn.Embedding module. In the function `train_data()`, it's written: ``` for i in xrange(max_sents):...

How to get nearest neighbours using the API?

4

Hi! I want to get the nearest neighbours inside my Java code using the JFastText API; however, I can not find which method I should use. In the original library...

LogLikelihood

2

Hi askerlee! I would ask you if the logLikelihood computed by "calcLoglikelihood" function is normalised by the number of words in the corpus? If not, it could be easily done?...

Sentences within a dataset

2

I am working on a dataset quite "noisy", so it's very difficult to exactly detect a sentence (for example, I have a lot of abbreviation with points, so these points...

Training on dataset without categories

3

Hi! I have a corpus of documents without categories, and I would generate topics without adding this kind of information. However, it seems that this code is only oriented to...

Https rather than http using OAuth protocol

1

The oauth_init() function needs to be updated with the address having "https" rather than "http", in the following way: ``` def oauth_init(self): """Start outh and return authorization url.""" service =...

Hardcoded pahts

1

Inside "paras.py" there are some paths hardcoded to your personal folder, generating an error when running the code. E.g.: `/shared/data/jiaming/local-embedding/sp` Just modified them to a local path the code works.

In run.sh, _word2ve**e**c_ should be _word2vec_: `gcc word2vec.c -o word2veec -lm -pthread -O2 -Wall -funroll-loops -Wno-unused-result`