1 of 21

Introduction to Speech & Natural Language Processing

Lecture 1

Information Layers of SNLP

Krishnendu Ghosh

2 of 21

Defining SNLP

Speech & Natural language processing (SNLP) is a field of artificial intelligence (AI) that allows machines to read, derive meaning from text, and produce documents.

​

It works in the background of many services, from chatbots through virtual assistants to social media tracking.

​

Such language technologies are already showing major penetration into the information and communication industry.

3 of 21

History of SNLP

1950-1960: prehistory, scientific knowledge regarding artificial intelligence and linguistics extremely limited

​

1960-1990: symbolic models, rules handwritten by experts, very limited coverage

​

1990-2010: statistical models, machine learning on data annotated by experts, good coverage

​

2010-present: neural models, machine learning on non-annotated data, excellent coverage

4 of 21

Stages of NLP

5 of 21

Stages of NLP: Phonetics

Phonetics is concerned with the production, transmission and perception of sounds, without prior knowledge of the language being spoken.

6 of 21

Importance: Phonetics

  • Foundation of Automatic Speech Recognition (ASR) and Text-to-Speech (TTS).
  • Helps map audio signals → phonemes → text.
  • Enables correct pronunciation modeling in multilingual systems.

​

Example:

​

Word

IPA (Phonetic)

Meaning

cat

kæt

Animal

bat

bæt

Animal

7 of 21

Stages of NLP: Phonology

Phonology studies the rules that organize patterns of sounds in human languages.

​

Example:

​

right/light, arrive/alive

​

Bengaluru

​

​

​

8 of 21

Importance: Phonology

  • Determines pronunciation rules (e.g., plural -s → /s/ vs /z/)
  • Important in speech synthesis and accent modeling.
  • Helps build phoneme-level models for low-resource languages.

​

Example:

​

Plural forms in English:

​

  • “cats” → /s/
  • “dogs” → /z/
  • “horses” → /ɪz/

​

9 of 21

Stages of NLP: Morphology

Morphology is the study of how words are composed by morphemes, which are the smallest meaningful units of language.

​

The structure of a word consists of several morphemes:

  • one root or stem
  • zero or more affixes, such as prefixes and suffixes

​

Example: draw, draw+s, draw+ing+s, un+draw+able

10 of 21

Importance: Morphology

  • Used in stemming, lemmatization, machine translation, search, and speech synthesis.
  • Reduces vocabulary size by grouping inflected forms.
  • Critical for morphologically rich languages.

​

Example:

​

unhappiness = un- (prefix) + happy (root) + -ness (suffix)

​

11 of 21

Stages of NLP: Syntax

Syntax studies the rules and constraints that govern how words can be organized into sentences.

​

Connection with formal language theory and rewriting grammars.

​

Differently from formal language theory, NLP systems need to be robust to input that does not follow the rules of grammar.

​

​

12 of 21

Stages of NLP: Phrase structure

Example:

​

Alice/N eats/V strawberries/N with/P chocolate/N

​

​

13 of 21

Stages of NLP: Dependency tree

Example : Rolls-Royce said it expects its U.S. sales to remain steady.

​

​

14 of 21

Importance: Syntax

  • Used in parsing, machine translation, and question answering.
  • Helps models understand sentence structure (subject, object, verb).

​

15 of 21

Stages of NLP: Semantics

Semantics is the study of the meaning of linguistic expressions such as words, phrases, and sentences.

​

Most common representations for semantic structure are:

  • logical representation
  • predicate-argument structure
  • graph representation

​

16 of 21

Importance: Semantics

  • Core for Word Embeddings (Word2Vec, GloVe), translation, chatbots, and question answering.
  • Helps resolve word sense disambiguation.

​

Example:

​

“bank” → [financial institution] or [river bank]

​

17 of 21

Stages of NLP: Pragmatics

Pragmatics studies the way linguistic expressions with their semantic meanings are used for specific communicative goals.

​

In contrast to semantics, pragmatics explicitly asks the question what an expression means in a given context.

​

An important concept in pragmatics is the speech act, which describes an action performed through language.

18 of 21

Importance: Pragmatics

  • Helps chatbots and dialogue systems infer intent, tone, and implied meaning.
  • Used in conversational AI and sarcasm/emotion detection.

​

Example:

​

“Can you pass the salt?” or “Can you open the window?” → not a question about ability, but a polite request.

19 of 21

Stages of NLP: Discourse analysis

Discourse analysis studies written and spoken language in relation to its social context.

​

Discourse refers to a piece of text with multiple sub-topics and coherence relations between them, such as:

  • explanation
  • elaboration
  • contrast

20 of 21

Importance: Discourse analysis

  • Used in text summarization, coreference resolution, and document understanding.
  • Captures relationships like cause-effect, elaboration, contrast.

​

Example:

​

“John bought a car. He loves it.” → “He” refers to “John”.

21 of 21