SEMANTICS IN COMPUTATIONAL LINGUISTICS

 SEMANTICS IN COMPUTATIONAL LINGUISTICS

1) What is Semantics?

Semantics is the study of meaning in language.
Other parts of linguistics study sounds (phonetics/phonology) or structure (syntax), words (morphology) but semantics asks:

·      What does a word or sentence mean?

·      How do we understand the idea behind the words?

Example:
“Rita opened the window.”
From this, we understand:

• Rita = the one who did the action
• opened = the action
• window = the thing affected

Semantics studies how these meanings are created.

2) Semantics Vs Syntax Vs Pragmatics

• Syntax = how words are arranged
• Semantics = what the words mean
• Pragmatics = meaning from context or intention

Example:
“Can you pass the salt?”

·        Semantic meaning: Are you able to pass it?

·        Pragmatic meaning: Please give me the salt.

3) Sentential Semantics: Meaning of Whole Sentences

Sentential semantics studies how entire sentences carry meaning. To be clear and precise, semantics often uses logic (because computers understand logic better than natural language).

Example:
“Pragnanadha played a Russian.”
This means:

·        There is a Russian person

·        Pragnanadha played that person (in chess)

Logical meaning helps computers answer questions like:

·        Who played whom?

·        Was the opponent Russian?

So, sentential semantics converts sentences into exact meanings.

Ambiguity in Sentences

Sentential semantics studies ambiguity (when a sentence has more than one meaning).

Example:
Visiting relatives can be boring.

Possible meanings:

1.   Relatives who visit can be boring.

2.   The act of visiting relatives can be boring.

4) Compositional / Lexical Semantics: Meaning from Word Parts

Compositional semantics says:

The meaning of a sentence comes from the meanings of its words + grammar.

Example:
“Every student danced.”

·        “student” = group

·        “every” = all

·        “danced” = action

Change one word and the meaning changes:

·        Some students danced

·        No student danced

·        Every teacher danced

This shows how small word changes cause big meaning changes.

 

5) Discourse Semantics: Meaning Across Sentences

Meaning does not come sentence by sentence only. It flows across paragraphs and conversations.

Example:
“A girl entered the room. She sat down.”
We know “she” = the girl.

Discourse semantics studies:

·        How pronouns refer back

·        How information continues

·        How meaning stays connected

Scientists use a model called DRT (Discourse Representation Theory) that works like a memory box storing people, objects, and events, so future sentences can refer to them.

How DRT works (Memory Box idea):

Sentence 1: A man entered a room.
→ DRT stores in its memory box:

·        Person: a man

·        Event: entering

·        Place: a room

Sentence 2: He sat on a chair.

·      The pronoun he looks into the memory box

·      Finds a man Place: a room

·      Correctly understands who “he” refers to

6) Event and Temporal Semantics: Meaning with Time and Actions

Many sentences describe events happening in time.

Example:
“Rahul kicked the ball yesterday.”
Event = kicked
Person = Rahul
Object = ball
Time = yesterday

Semantic analysis must capture all of these.

Time meanings (Tense)

·        She sings → now

·        She sang → past

·        She will sing → future

·        She had sung → completed earlier

Temporal semantics helps in understanding stories, news, and sequences of events.

 

7) Deep vs Shallow Semantics

Computational linguistics uses two types of meaning analysis.

Deep Semantics

·        Uses logic and detailed meaning

·        Very accurate

·        Slow and complex
Like reading a text carefully.

Shallow Semantics

·        Uses keyword matching or statistics

·        Fast

·        Not always accurate
Like quickly skimming a text.

Example:
“I don’t like chess anymore.”
A shallow system sees “like” → thinks it’s positive.
A deep system sees “don’t” + “anymore” → understands dislike.

Modern AI combines both for better performance.

 

Conclusion

Semantics is the study of meaning. It helps us understand:

• what words refer to
• who does what
• when actions happen
• how sentences connect
• how meaning changes with context

Semantics is essential for:

• translation
• chatbots
• Siri/Google Assistant
• question answering
• text understanding

Without semantics, computers can talk, but they can’t understand.