0.25
0.5
0.75
1.25
1.5
1.75
2
Towards Universal Paraphrastic Sentence Embeddings
Published on May 27, 20162330 Views
We consider the problem of learning general-purpose, paraphrastic sentence embeddings based on supervision from the Paraphrase Database (Ganitkevitch et al., 2013). We compare six compositional archit
Related categories
Chapter list
Towards Universal Paraphrastic Sentence Embeddings00:00
Goal - 100:10
Goal - 200:52
Where do we start?01:41
The Paraphrase Database02:36
Modelling composition03:18
Objective function - 103:35
Objective function - 204:16
Evaluation04:29
In-domain datasets05:18
Average Pearson’s correlation on 22 datasets - 105:42
Average Pearson’s correlation on 22 datasets - 207:12
Scaling up08:02
Reflection08:26
Length08:59
Overfitting on in-domain data09:19
Parameter tuning - 109:34
Parameter tuning - 209:47
Other use cases?09:56
Initialization/Regularization - 110:43
Initialization/Regularization - 211:20
Qualitative/Quantitative analysis11:52
Character n-gram model - 114:03
Character n-gram model - 214:57
Character n-gram model - 316:14
Conclusion16:29
Thank You17:18