Semanlink - Embeddings AND Survey / Review

- "facts only" based embeddings
    - Translational Distance Models, distance-based scoring functions
       - TransE and co: entities and relations as points in vector space
           - TransH: relation-specific hyperplanes.
       - Gaussian Embeddings: entities and relations as random variables
    - Semantic matching models, similarity-based scoring functions
        - RESCAL (entities as a vector, relation as a matrix) and familiy
            - DistMult (relation as diagonal matrix)
            - Holographic Embeddings (HolE)
            - etc.
        - Matching with Neural Networks
            - SME, etc
- incorporating other info: entity types, relation paths,
textual descriptions, as well as logical rules
    - entity types
        - semantically smooth embedding
(SSE):  entities of the same type to stay close
to each other in the embedding space
        - type-embodied knowledge representation
learning (TKRL), which can handle hierarchical
entity categories and multiple category labels
- Applications in downstream tasks

2019-05-27 About

On sentence representations, pt. 1: what can you fit into a single #$!%@*&% blog post? · Supernatural Language Processing

Tags:

2018-11-27 About

what are the pros and cons of the various unsupervised word and sentence/ document embedding models? - Quora

Tags:

2018-08-19 About

Evaluation of sentence embeddings in downstream and linguistic probing tasks

Tags:

2018-06-27 About

[1805.04032] From Word to Sense Embeddings: A Survey on Vector Representations of Meaning

Tags: