Topic 01 — Module 1
Semantic vs
Lexical Search
Ye samajhna important hai ke aaj SEO mein "Semantic" itna kyun sunne ko milta hai — aur purani keyword pe based approach kahan kamzor hai.
Lexical Search ek lafz pe based search system hai. Iska simple matlab: search engine sirf dekhhta hai ke user ne jo word type kiya, woh kisi page ke URL, title ya content mein likha hua hai ya nahi.
Agar koi "best dog foods" search kare, toh lexical search engine har page scan karega:
best dog foods
Jo page mein yeh phrase jitni zyada baar aaya, algorithm usse zyada relevant samjhega. Meaning nahi samajhta — sirf words count karta hai.
Google ne khud isko "String Matching" kaha tha. Lexical = literal characters match. Agar page ne "side effects of dog food" likha tha lekin "best dog foods" bhi mention tha, toh lexical algorithm us page ko pick kar sakta hai — chahe wo actually relevant na ho.
📖 Lexical Search — Definition
Search engine information objects (web pages, documents) mein literal words ya strings match karta hai taake user query fulfill ho sake. Meaning, context ya intent ko ignore karta hai.
Lexical Search ke Algorithms
1
Bag of Words Model
Page ko words ka ek bag samjha jata hai. Har word ki frequency count hoti hai bina kisi order ya context ke.
2
TF-IDF (Term Frequency — Inverse Document Frequency)
Ek specific algorithm jo count karta hai ke ek term ek document mein kitni baar use hua. Zyada frequency = zyada significance.
3
BM25 Language Model
Standard retrieval model jo TF-IDF ka improved version hai. Abhi bhi kuch systems mein use hota hai.
⚠️ Lexical Search ki Limitations
- ❌ Meaning ya intent nahi samajhta
- ❌ Entities, unke attributes aur context ko completely ignore karta hai
- ❌ Sirf literal characters/strings pe focus — real-world concepts nahi
- ❌ "Things" ko nahi samajhta, sirf "Strings" dekhta hai
"Things, not Strings"
— Amit Singhal, Google Engineer · May 16, 2012 · Knowledge Graph Launch Post
16 May 2012 ko Google engineer Amit Singhal ne ek historic blog post publish ki jisme unhone announce kiya: Google ab entities ko samajhne lagega.
Unhone bataya ke jab aap "Taj Mahal" type karke search karte hain, toh Google ab samajhta hai ke user ka kya irada hai — kyunki wo context samajhta hai. "Things" yaani entities — aur "Strings" yaani keywords. Lexical matching band, entity recognition shuru.
💡 SEO Practical Implication
Jab aap apne content mein sirf keywords match karte hain — toh aap purani lexical approach follow kar rahe hain. Modern SEO mein aapko
entities, their attributes aur context ko address karna padta hai.
Semantic Search ek broader search paradigm hai jisme multiple methods use hote hain sirf isliye ke context, user intent aur entities samjhe ja sakein.
Lexical Search
String-Based
- Literal word match karta hai
- Frequency count pe rely karta hai
- Meaning ignore karta hai
- Context nahi samajhta
- Entities ko nahi pehchanta
Semantic Search
Meaning-Based
- Contextual meaning process karta hai
- Entities identify karta hai
- User intent samajhta hai
- Context-aware results deta hai
- Information relevance check karta hai
Semantic search engine ka primary goal hai: information need accurately anticipate karna aur address karna. Pehle query samjho, phir relevant documents dhundo.
Jab user ne "best dog foods" search kiya, toh search engine sabse pehle us query ko break down karta hai:
1
Query Parsing (Milliseconds mein)
Search engine query ko samajhta hai. Entities identify karta hai: Dog (entity) + Food (entity) + Best (qualifier — user list chahta hai).
2
Intent Identification
User ka actual goal kya hai? "Best" word se pata chalta hai ke user ek comparison ya listicle chahta hai.
3
Query Processing — Documents Dhundna
Jab query fully samajh aa jaaye, tab algorithm start hota hai aur relevant documents retrieve karta hai. Yahan hamari website ki game shuru hoti hai.
🔑 Key Insight
Query Parsing + Query Processing — ye do stages hain. Pehle samajhna, phir dhundna. Ye sab nanoseconds mein hota hai.
⭐ PAA Section — Content Strategy ka Khazana
"People Also Ask" aur "People Also Search For" — ye dono sections Google ne khud recommend kiye hain. Ye sections show karte hain ke ek topic ke related kya aur queries hain.
Ye aapko batata hai:
- 📌 Related topics kya hain
- 📌 User journey mein next/previous step ki search queries
- 📌 Kya ye same page mein cover hoga ya alag page banana padega
Search: best dog foods
PAA suggestions (har ek ek alag intent reveal karta hai):
- → What is the healthiest food for a dog?
- → Which dog puze best in Pakistan? (localized)
- → Can dogs eat roti? (regional context)
Har ek PAA suggestion ek different contextual layer hai — ek alag user intent. Semantic search engine ne "best" ke multiple possible intents ko proactively augment kiya.
Local SEO mein yeh difference sabse zyada practically samajh aata hai. Jab koi Lahore mein "best dentist near me" search karta hai — lexical engine sirf woh pages dhundhta jis mein "best dentist near me" literally likha ho. Magar semantic engine kuch aur karta hai:
Search: best dentist near me (Lahore se)
1
Entity Parsing
Dentist = Healthcare professional entity. "Near me" = location intent signal — user apni current location ke paas dhundh raha hai.
2
Contextual Enrichment
Search engine user ki location (Lahore, DHA Phase 5) identify karta hai. Ab "dentist" + "Lahore" + "DHA" ek rich context ban jaata hai.
3
Intent Fulfillment
Google Maps results, reviews, opening hours, contact number — yeh sab ek local entity ke attributes hain jo search engine surface karta hai.
💡 Local SEO Practical Implication
Agar aapka dental clinic sirf apni website pe "dental services" likh ke baitha hai bina
NAP (Name, Address, Phone), Google Business Profile, aur local entity signals ke — toh lexical approach zyada kaam nahi karti. Semantic SEO mein local entity ko
properly define karna padta hai — exact address, area name, landmark ke references, aur local citations.
📍 Local Semantic Signals — Checklist
- 📌 Google Business Profile — poori complete entity definition
- 📌 NAP (Name, Address, Phone) — har jagah consistent
- 📌 Neighborhood mentions — "DHA Lahore," "Gulberg," "F-7 Islamabad"
- 📌 Local landmarks as contextual anchors — "Packages Mall ke qareeb"
- 📌 Schema.org LocalBusiness markup — entity ko structured format mein define karo
Ecommerce mein lexical approach ka matlab hai: product title mein keyword thosa, category page mein baar baar repeat karo. Semantic approach bilkul alag hai — product entities properly define karo.
Daraz / Amazon example:
Ek user ne search kiya:
wireless headphones under 5000 for gaming
Lexical engine: Pages dhundhega jisme ye exact phrase hai.
Semantic engine kya samajhta hai:
- → Entity: Wireless headphones (product category)
- → Attribute: Price constraint — under PKR 5,000
- → Attribute: Use case — Gaming (low latency, mic, surround sound important)
- → Intent: Transactional — user khareedna chahta hai
🛒 Ecommerce SEO Shift
Ek ecommerce product page jo sirf keyword-stuffed hai woh outdated approach hai. Modern semantic approach mein:
product entity ke sab attributes clearly define karo — specs, use cases, compatibility, price range, brand entity, aur customer intent signals. Yahi Google ko quality content lagti hai.
Semantic search engine sirf aaj ki query nahi dekhta — wo user ki search history bhi note karta hai.
📋 Google Patent: Context Sensitive Ranking (2029 expire)
Ye patent batata hai ke Google kaise ek searcher ki history apne paas save karta hai:
- • Aapne kya search kiya tha
- • Kis result pe click kiya
- • Kitni der us website pe rahe
- • Kaun se internal links click kiye
- • Agar page se bounce kiya toh — signal: user satisfied nahi hua
Agar aapne ek page dekha, thoda jhanka, aur band kar diya — search engine ne note kar liya: click satisfaction nahi hui. Phir se search bar pe aana = user aur dhundh raha hai.
2019 mein Google ne BERT (Bidirectional Encoder Representations from Transformers) launch kiya. Yeh ek AI model hai jo har query ko poore context mein samajhta hai — sirf words ek-ek karke nahi, balke puri sentence ka meaning ek saath.
BERT se pehle vs baad — ek example:
Search (2018 se pehle)
"brazil traveler to usa need visa"
Old engine: sirf "visa" + "USA" match karta — US citizens ke liye Brazil visa results deta tha
BERT ke baad (2019+)
"brazil traveler to usa need visa"
BERT samajhta hai: "to" ka matlab direction — Brazil se USA jaane wala banda, us ke liye visa info chahiye
2021 mein Google ne MUM (Multitask Unified Model) launch kiya — BERT se 1000x zyada powerful. MUM ek saath multiple languages mein samajhta hai aur ek complex question ka jawab multiple formats (text, images, video) se de sakta hai.
🤖 SEO Implication — AI Era
BERT aur MUM ka matlab hai ke aapki content ka
actual meaning matter karta hai — sirf keywords nahi. Aapka content jo insaan padhte hain woh naturally meaningful hona chahiye. Keyword stuffing ka time gaya.
Pakistan mein Semantic SEO ki importance samajhne ke liye hum kuch aise examples dekhte hain jo aap roz ki zindagi mein use karte hain.
Example 1 — Daraz.pk Product Search
Samsung mobile under 40000 with good camera
❌ Lexical: Sirf pages dhundhta jisme "Samsung mobile under 40000 with good camera" literally likha ho
✓ Semantic: Samajhta hai — User chahta hai: Samsung phone + price < 40k PKR + camera quality high hona chahiye (megapixels, aperture, AI camera features). Intent: transactional purchase.
Example 2 — News Search (Dawn, Geo, ARY)
Imran Khan case latest update today
Semantic Engine samajhta hai: "Imran Khan" ek person entity hai (Pakistan ka siyasatdan). "Case" = legal matter. "Latest update today" = recency signal — user abhi ki news chahta hai. Toh search engine fresh news articles prioritize karta hai.
Example 3 — Local Business Search
biryani Karachi DHA delivery
Semantic Engine: "Biryani" = food entity (Pakistani cuisine). "Karachi DHA" = geographic entity. "Delivery" = service attribute + user intent (ghar baith ke mangwana chahta hai). Google Maps results + Food delivery apps surface karta hai automatically.
🇵🇰 Pakistan Specific SEO Tip
Pakistan mein local businesses ke liye
Urdu + English bilingual entity signals important hain. Agar aapka restaurant "DHA Lahore" aur "ڈی ایچ اے لاہور" dono mein mention ho — semantic engine ko zyada confidence milti hai aapki entity recognize karne mein.
Q1
Lexical search aur Semantic search mein sabse bada difference kya hai?
Answer: Lexical search sirf words/strings match karta hai bina meaning samjhe. Semantic search context, intent aur entities samajhta hai.
Q2
Google ka "Things not Strings" kab launch hua aur kisne kaha?
Answer: 16 May 2012 ko Google engineer Amit Singhal ne Knowledge Graph launch ke saath yeh phrase use kiya.
Q3
TF-IDF ka pura naam kya hai aur yeh kya karta hai?
Answer: Term Frequency — Inverse Document Frequency. Yeh count karta hai ke ek word document mein kitni baar aaya — zyada frequency matlab zyada relevance (lexical approach).