Simeng Han

icon company Yale University icon location NLP Research

Results 7 comments of Simeng Han

TextRank for CNN/DailyMail

Got it. Thanks!

Depth loss

Hi @ozansener @SimonVandenhende , The disparity data read from the Cityscapes dataset needs to be first processed to be converted to depth: `depth = baseline * fx / disparity, `...

Depth loss

Hi @ozansener , I just found that in the table of the paper you reported "1 / Disparity_Error", so is it Disparity instead of Depth that you are reporting? I...

Depth loss

Thanks for the clarification!

Unable to generate the JSON file from the excel file

try this [https://github.com/stevetarver/excel-as-json/issues/46](url)

Why tokenizing 2 times ?

It seems that Stanford CoreNLP is only for tokenizing into sentences while BertTokenizer is for tokenizing into subwords to produce the vocabulary. https://github.com/nlpyang/PreSumm/blob/ba17e95de8cde9d5ddaeeba01df7cace584511b2/src/prepro/data_builder.py#L123

Translation of 'Some As are Bs' sentences

Yes, we are aware that conjunction should be used with the existential quantifier. We corrected most of the errors during the review stage where implication was used with the existential...