Everything about semantic analysis in rnn

What is semantic analysis?

The main work of semantic analysis is to understand the meaning of a sentence. So we can say that semantic analysis means giving the exact meaning of the text.

Types of semantic analysis:

Lexical semantic:
It means, the meaning of an individual word.
Suppose we have a sentence. So what we will do is that, simply we will find the meaning of each word , and after getting the meaning of the word we add the meanings one after another to create a sentence.

Compositional semantic or combination of work:
It involves how a word combines a larger meaning or sentence. Here the meaning of each word matters. But syntax and the sentence construction play an important role
Here we find:
1. Relationship exist words
2. The word order
3. Context of sentence
4. Semantic structure
5. Real word knowledge

After getting all this information we get meaning of a sentence

Let's take two sentence for example:
1. I love you
2. You love me
Here both sentence meaning of I, me and you is same but the meanings are not same.
In first sentence I = subject, you = object
But,
In second sentence you = Subject and me = object.

So here we saw that same word but containing different meanings when used in a different sentence.
This is called the principle of compositionality

Using leximal semantic + composition semantic = semantic analysis
So we can say that using leximal semantic and composition semantic we do semantic analysis means get a sentence meaning.

Basic Building Blocks of Semantic analysis:

Entities:
This represent individuals like location, person ,etc.
Example: Dhaka, Joe Biden

Concept:
It represent general category of individuals like animals, fruits, person, etc.

Relations:
Here we find the relation between entities and concept.
Example: Apple is a fruit

Predicates
It represent verb structure.

Case grammar: Here we analyze the structure of sentence

How semantic analysis is performed?

There are lot of ways. Let's discuss three ways:

First order predicate logic:

This technique converts sentence into a logical form.
Example 1:
Rafsun loves car
Logical form = loves(Rafsun, car)

Example 2:
Rafsun takes Engineering and Pharmacy
Logical form = takes(Rafsun, Engineering) Vtakes(Rafsun, Pharmacy)

Semantic Nets

This method shows the semantic relationship between concepts. Here concept means general category.

Here in the image we can see the relation between general category like a bird is a pet or fish lives in water, etc. In this way getting the meaning of a sentence becomes more easier.

Case Grammar:

Here we analyze the structure or the sentence

In the image can see a sentence: We gave our cat a food.
Here,
We = subject, gave = verb, dog = indirect object, food = direct object.
What we gave? food and a is recipient. So we can understant the structure of the sentence this way in this method.