It learns vector representations of individual sentences and then calculate the semantic relationship between sentences based on vector distance.

University of Michigan - Ann Arbor

1. Sentence Encoding Models
2. Sentence Pair Interaction Models 

Sentence Representation Model Categorizations

Neural Network Models for Paraphrase Identification, Semantic Textual Similarity, Natural Language Inference, and Question Answering

Sentence Representation Model

It uses some sorts of word alignment mechanisms (e.g., attention) then aggregate inter-sentence interactions.

Sentence Pair Interaction Model

1.  Incorporating inter-sentence interactions is critical.
2. Tree-LSTM does not help as much as previously claimed but surprisingly improves performance on Twitter data.
3. Enhanced Sequential Inference Model has the most consistent high performance for larger datasets, while Pairwise Word Interaction Model performs better on smaller datasets and Shortcut-Stacked Sentence Encoder Model is the best performaning model on the Quora corpus.

Some NN Design Hints for Sentence Representation

1. The input embedding layer.
2. The context encoding layer.
3. The interaction and attention layer.
4. The output classification layer.

Learn Before

Related