NLP Paper Archive aims to be a sort of library containing interesting paper talking about Natural Processing. It's organised in topics (i.e. word embedding, ethical application, ecc.) and every paper it's with a description made by important keywords, main questions around which the paper is focused and main ideas proposed.
Distributional Hypotesis, Harris' Distributionalism, Structuralism, Matrix Models, Latent Space
- How is it possible that a latent discriminating space underlying the relations between terms and contexts in a given raw record of linguistic performances succeeds in capturing a surprising amount of linguistic properties, including syntactic regularities and structures, semantic relatedness, isomorphic configurations between different languages and historical change? (Chapter 4)
- But what exactly is a context? And what is its precise role in the production of linguistic meaning? (Chapter 4)
- Distributional Hypothesis: You shall know how word by the company it keeps. The degree of semantic similarity between two linguistic expressions A and B is a function of the similarity of the linguistic contexts in which A and B can appear. (Chapter 4)
- “The meaning of a word is defined by the circumstances of its use”(Manning and Schutze, 1999) (Chapter 4)
- Harris' Distributionalism: the distributional chratcter of meaning is perfectly indipendendent from meaning. Harris’s distributionalism conveys an image of language as an autonomous system of regularities and constraints, not governed byany external source of meaning.
- Structuralist Hypothesis: the meaning is the effect of the structure.