Semantics
In this blog, I’m going to explain semantics. But first, I have an important question?
What is a word?
As a programmer, a word is string of a sequence of characters.
Well, what about the meaning of this term in linguistics?
Maybe we can define the term ‘word’ as a set of characters that define a specific meaning such as a concept or idea which we give it a term in linguistic sciences that the word term has meaningful features which is called semantic.
Semantic can be inform of relation between words that can be understood from the context. If we think about words like water and H20, or coach and sofa, or big and large. These set of words have a relation of synonymity. They are synonyms, however they are not perfect synonyms. Meaning that not both words can fit into a text. The best fit will always depend on the context. Embeddings through word2vec or doc2vec are trained in a large amount of vocabularies using different features or dimension which enable it to predict the best word fit into the text.
Antonymy is also a strong relation between words where one word is an opposite to the meaning of another word. Meanings such as wet/dry, cold/hot, happy/sad …etc. Words similarity is another relation between words such the county and city, or car and bus, or lamb and goat. We can see that these words have kind of similar relation in meaning for example we can know that car and bike are both have transportation relation, while country and city are both represent geographical dimension, however goat and lamb have a cattle relation. Based on that, we have an important term in linguistic science i.e., called lexicons where we can determine how much words are similar to each other.
What is Lexicon?
Lexicon is a list of words that includes annotation of the similarity score ranging from. -1 to +1. If the score is. close to -1, this means the word is negative similarity, but if it is zero this means the word is neutral word and if it is close to +1 , it is positive similarity.
The fourth word relation can be found among semantics is word association.
What is Word Association?
Word Association focuses on the dependency relation between words. In the previous example we saw that car/bike both share common relation that they are means of transportation. However the set of car/gas shares the relation of dependency.
Semantic Domain
Semantic domain covers vocabulary that is related to a specific field. It shows the structural relation between words. House can be domain because it is a basic structure that includes roof, doors, windows, rooms, garage..etc. These words are related to each other and comes under a basic semantic field. The semantic structure can be divided into superordinate and subordinate.
Superordinate and Subordinate Semantic
Superordinate concept reflects the word as a main semantic, while subordinate word is a subclass of a main structure. If we think of the word ‘vehicle’ and other related words such as car, truck, or bus.
We can recognize that vehicle is superordinate word while car, truck and bus are lower semantic because they are under the category of vehicle.
Affective Lexicon
Affective lexicon is an important concept in NLP. Words which reflect an affective meaning either as an emotion which is called connotation or as an evaluation. It can be positive connotation such as happy or negative connotation as sad. It can also be positive evaluation as great, or good job or as a negative evaluation as bad, or miserable.
Let’s wrap up today’s post. We described the concept of the word and introduce the term of semantic where we explain the relations between words. These relations can be one or more of the following (synonymy, antonymy, similarity, dependency or relatedness, superordinate/subordinate, connotation).
References:
1-Dan Jurafsky, Speech and Language Processing, Chapter 6,Vector Semantics.
2- Motaz Saad,NLP Vectors & Semantics and Word Embeddings.
