Sentence Embedding Model

A Sentence Embedding Model is a text-item embedding model for sentences that can be referenced by a sentence embedding encoder.

• AKA: Sentence Encoding Model.
• Context:
  • It can (often) be based on neural network architectures
  • It can (often) be trained on a Large Corpus of Text.
  • It can utilize techniques such as Word Embedding Aggregation, Sequence Modeling, and Attention Mechanisms.
  • It can range from being a Pure Sentence Embedding Model to being a Contextually-Informed Sentence Embedding Model.
  • It can range form being a Domain-Specific Sentence Embedding Model to being a Universal Sentence Embedding Model.
  • ...
• Example(s):
  • one based on:
    • S-BERT, which generates context-sensitive embeddings for each word in a sentence and can be adapted to generate sentence embeddings.
    • GloVe (Global Vectors for Word Representation) combined with average pooling, where word embeddings are averaged to produce a sentence embedding.
  • ...
• Counter-Example(s):
  • A Word Embedding Model, which focuses on generating embeddings for individual words rather than whole sentences.
  • A Document Embedding Model, which is designed to generate embeddings for entire documents, which may contain multiple sentences or paragraphs.
• See: Vector Space Model, Semantic Analysis, Word Embedding, Document Embedding.

---

---