In NSM, semantic primes are the most fundimental “lexical units” (so they can be words, or morphemes, etc. the size doesn’t matter) across languages.

They are the “core of a universal mental lexicon”.

There are…

guidelines for identifying semantic primes

1. A semantic prime has to be found in every(ish?) natural language
2. A semantic prime has to be indefinable by other primes

proof for the existence of semantic primes

Proof: given if the Strong Lexicalization Hypothesis holds, semantic primes must exist.

SVF is a standardized Discourse-Completion Task for verbal recall and fluency. It is administered by asking the participant to recall a bunch of words from within a category within 60 seconds.

The semiconductor industry is a growing industry, the beginning of the semiconductor industry was actually in the silicon valley.

We are now taking a look at a reticle.

algorithms used in the semiconductor industry

Per KLA —

Classification

• Random forest
• Boosted decision trees
• MLPs
• CNNs

Reference generation

• GANs (WAT)
• VAEs

Natural Grouping and Clustering

• auto-encoders
• manual feature extractors

sense is the meaning of the word.

synonymy

synonymy—using synonyms as a proxy for word meaning.

But! There are probably no examples of perfect synonymy: two synonyms have slightly different sense (“my big sister” != “my large sister”)

word relatedness

not synonyms, but closeness in utility or semantic frame:

• coffee + tea — similar
• coffee + cup — related, but not similar

semantic field

words that relate by covering a similar semantic domain:

Common algorithm

1. tokenization
2. classier for whether a period token is a part of a word or the sentence boundary