search

facebookresearch / SentEval

A python tool for evaluating the quality of sentence embeddings.
Other
2.09k stars 309 forks source link

What's the purpose of SentEval and how can i integrate it it with my inferSent model? #54

Open prachij94 opened 5 years ago

prachij94 commented 5 years ago

Hi,

Can you please elaborate how can i use SentEval integrated with my inferSent model to get the sentence embeddings for my own sentences corpus?

It will also be great to get more insights on interpreting and using the results for similar sentences for our business requirement(sentiment analysis for the keywords entered) which are as follows for the STS16 transfer task used:

{
'STS16':{
'answer-answer':{
'pearson':(0.25246563397263544, 4.702985203895257e-05 ), 'spearman':SpearmanrResult(correlation=0.3079584675460508, pvalue=5.556290983707961e-07), 'nsamples':254 }, 'headlines':{
'pearson':(0.3635692171614175, 3.385973583781225e-09 ), 'spearman':SpearmanrResult(correlation=0.3514220638099597, pvalue=1.1960065660420662e-08), 'nsamples':249 }, 'plagiarism':{
'pearson':(0.4232950489009525, 2.050858700377934e-11 ), 'spearman':SpearmanrResult(correlation=0.4132543728895399, pvalue=6.698065484617935e-11), 'nsamples':230 }, 'postediting':{
'pearson':(0.6907186645066123, 6.277476490673788e-36 ), 'spearman':SpearmanrResult(correlation=0.7436531853554167, pvalue=3.323825847627377e-44), 'nsamples':244 }, 'question-question':{
'pearson':(-0.028058870107370452, 0.6867356266558649 ), 'spearman':SpearmanrResult(correlation=-0.042384317092789704, pvalue=0.5422977635039196), 'nsamples':209 }, 'all':{
'pearson':{
'mean':0.3403979388868495, 'wmean':0.34964917170036636 }, 'spearman':{
'mean':0.35478075450163543, 'wmean':0.3654022810828834 } } } }

JaSchuste commented 5 years ago

I got the same question right now. Did you figure it out and maybe tell me how this is possible?