Does Siksha Sarovar have an AI chatbot to answer student doubts?

Yes. Siksha Sarovar has a built-in AI Assistant chatbot accessible from a floating button on every page. It understands English, Hindi and Hinglish, handles typos (for example 'pyhtion' or 'certifecate'), and indexes 165+ destinations including every course, lesson, BCA subject, school chapter, competitive exam topic, FAQ and tool. Most queries return direct link cards in under 5 milliseconds. An AI fallback is available for novel questions.

Can I ask the SikshaSarovar chatbot questions in Hindi or Hinglish?

Absolutely. The chatbot is built specifically for Indian students — natural Hinglish queries like 'kaise milega certificate', 'free hai kya', 'pyhtion ke datatype kaha hai', 'kaha se shuru karu' are first-class citizens. The matcher strips Hindi filler words and routes you to the right course, lesson or page.

Is the SikshaSarovar AI chatbot free to use?

Yes. The chatbot is 100% free, requires no signup, and is available on every page. It runs locally in your browser for the vast majority of queries — there is no API cost or usage limit. The optional 'Ask AI' fallback for advanced coding questions uses the Pro AI Tutor.

Is Siksha Sarovar really free?

Yes. Every course, lesson, quiz, online compiler, and notes download is free to use without an account. We offer an optional Pro pass that unlocks longer AI tutor sessions, larger compiler quotas and priority support, but it is not required to learn from the platform. The educational content itself stays free.

Do I need to sign in to use the courses?

No. You can browse any course, read all lessons, run code in the compiler and take quizzes without signing in. Google Sign-In is purely optional and is used only to save your progress, quiz scores and certificate eligibility across devices. We never request access to Gmail, Drive, Calendar, Contacts, or any sensitive Google data.

Are the certificates from Siksha Sarovar recognised?

Our certificates are a record of completion that you can share on LinkedIn or attach to applications, but Siksha Sarovar is an independent platform — not a UGC-recognised university or board. We are upfront about that. The certificate is most useful as a verifiable signal that you have completed the curriculum, not as a substitute for a degree.

Which courses are best for BCA and MCA students?

Our University Curriculum section covers the YMCA BCA/MCA syllabus subject-by-subject — Data Structures, DBMS, Web Based Programming, Computer Networks, Operating Systems, Software Engineering, Data Warehousing and more. Each subject is broken down into the same units your university teaches, with previous year question papers where available.

Can I use Siksha Sarovar to prepare for SSC, UPSC, Banking or Railway exams?

Yes. The Competitive section has dedicated tracks for SSC (CGL, CHSL, MTS), UPSC, IBPS/SBI Banking, RRB Railways and defence exams (NDA, CDS, AFCAT). Topics include quantitative aptitude, reasoning, English grammar, general knowledge and current affairs, written specifically for the Indian exam pattern.

What languages does the online compiler support?

The Siksha Sarovar online compiler supports C, C++, Python, Java, PHP, JavaScript, C# and SQL. The compiler runs your code in a sandboxed environment using Judge0, returns the standard output and error stream, and supports stdin so you can test interactive programs. There is no installation — everything runs in your browser.

How is my personal data handled by Siksha Sarovar?

We follow data minimisation: we collect only what is needed (email, name, profile picture from Google sign-in, and your learning progress). Data is stored on Supabase with HTTPS in transit. We do not sell user data, and we do not use it to train AI models. You can request deletion at any time by emailing contact@sikshasarovar.com — see our Privacy Policy for the full details.

Who founded Siksha Sarovar?

Siksha Sarovar was founded by Rohit Kumar, who serves as CEO and Head Developer. Rohit built the platform to provide free, structured education to students across India — covering programming courses, university notes, school study material and competitive exam preparation.

Unit 2 — Named Entity Recognition (NER): Concepts & Techniques — Natural Language Processing Notes

Named Entity Recognition (NER)

Named Entity Recognition (NER) identifies and classifies "named entities" in text into predefined categories — people, organizations, locations, dates, monetary values, and more.

"Riya joined Google in Bengaluru on 5 August 2026 for ₹18,00,000."
   [PERSON]    [ORG]    [LOC]         [DATE]              [MONEY]

Common Entity Types

Type	Examples
PERSON	Riya, Elon Musk
ORG (Organization)	Google, IIT Delhi
GPE / LOC (Geo-political entity / Location)	Delhi, India, Mount Everest
DATE / TIME	5 August 2026, 10 AM
MONEY	₹18,00,000, $500
PERCENT	25%
PRODUCT	iPhone 17

Why NER Is Hard

Ambiguity — "Washington" could be a person (George Washington), a place (Washington D.C.), or an organization.
Boundary detection — is "New York Stock Exchange" one entity or three?
Novel/unseen entities — new company/product names appear constantly; a dictionary-only approach fails.
Case sensitivity issues — many NER systems rely on capitalization, which fails on lowercase text (social media, ASR transcripts).

NER as a Sequence Labeling Problem — BIO Tagging

NER is typically framed the same way as POS tagging (previous lessons): assign a label to every token, using the BIO scheme:

Prefix	Meaning
B-`TYPE`	Beginning of an entity of that type
I-`TYPE`	Inside (continuation) of an entity
O	Outside — not part of any entity

Token:  Riya    joined  Google  in   Bengaluru
Tag:    B-PER   O       B-ORG   O    B-LOC

Token:  New     York    Stock   Exchange  opened
Tag:    B-ORG   I-ORG   I-ORG   I-ORG     O

NER Approaches

Approach	How it works	Example
Rule-based / Gazetteer	Dictionary lookup + hand-written regex patterns	List of all known country names
Statistical (HMM/CRF)	Sequence models trained on tagged corpora, similar to POS tagging's HMM (previous lesson)	Conditional Random Fields (CRF)
Neural / Transformer-based	Deep contextual models (BiLSTM-CRF, BERT-based)	spaCy's default pipeline, BERT-NER

CRF (Conditional Random Field) — Brief Intuition

Where an HMM models P(word | tag) and P(tag | previous tag) separately (generative), a CRF directly models P(tag sequence | word sequence) (discriminative) and can use arbitrary overlapping features of the input (capitalization, prefixes/suffixes, surrounding words) — which is why CRFs historically outperformed HMMs for NER.

Useful CRF features for NER:
  - Is the word capitalized?
  - Is the word in a gazetteer (known-entity list)?
  - What is the previous/next word?
  - What is the word's POS tag?
  - Does the word contain digits (helps detect DATE, MONEY)?

We put NER into practice with a production-grade library in the next lesson.