Yaya inverted index yake aiki?

Question

IT Interview Practice · Accepted Answer

**Inverted index yana haɗa term da documents da suke ɗauke da shi.** Hakan yana barin search engine ya gano candidates ba tare da karanta duk document daga farko zuwa ƙarshe a kowane search ba.

Ka tuna index a bayan littafi: nemi kalma, sannan ka je shafukan da aka lissafa. A search engine, document identifiers suna maye gurbin lambobin shafi.

## 1. Fara da ƙaramin collection

A ce muna da documents uku da analyzer mai sauƙi wanda yake mayar da rubutu lowercase kuma ya raba kalmomi. Analyzer shi ne jerin matakai da suke mayar da text zuwa terms da ake iya nema.

```text
D1: Red apple
D2: Green apple
D3: Red car

apple -> D1, D2
car   -> D3
green -> D2
red   -> D1, D3
```

Asalin hanya daga document zuwa words ce. Index yana juya ta zuwa word zuwa documents, shi ya sa aka kira shi inverted. Jerin documents da suke dacewa da term shi ake kira posting list.

## 2. Yi amfani da lists don amsa query

Ga query da yake buƙatar red da apple tare, karanta posting lists ɗin kuma ka ɗauki intersection: D1 yana cikin duka. Idan ɗaya daga cikin terms ya isa, ɗauki union: D1, D2 da D3 su ne candidates.

Misalin yana bayyana matching, ba ranking ba. Idan documents da yawa sun dace, engine zai iya ba su score bisa term frequency da yawan term ɗin a cikin collection. Gano matches da tsara matsayinsu matakai ne masu alaƙa amma daban.

## 3. Sanya indexing da search su dace

Dole a yi analysis na query ta hanyar da ta dace da indexed terms. Idan indexing ya mayar da Red zuwa red amma search yana neman literal Red, ana iya rasa match da ake so.

Analysis mai kyau ya dogara da field. Product descriptions suna amfana da text analysis; exact order identifier yawanci yana buƙatar exact matching. Raba identifier zuwa kalmomi na yau da kullum na iya sauya ma’anarsa. Duba bayanin Elastic na [full-text search](https://www.elastic.co/docs/solutions/search/full-text/how-full-text-works).

## 4. Adana isasshen bayani ga tambayar da ake yi

Document mai red da apple ba lallai ya ƙunshi phrase red apple ba. Don phrase queries, index yana buƙatar word positions don engine ya duba order da adjacency.

- Document identifiers suna nuna documents da suke ɗauke da term.
- Frequencies na taimakawa bayyana sau nawa term ya bayyana.
- Positions suna amsa inda terms suke dangane da juna.

Adana ƙarin bayani yana ba da queries masu zurfi amma yana ƙara index size da indexing work. Searches da application yake buƙata su ne suke tantance index mai amfani.

## 5. Fahimci kuɗin riƙe shi sabo

Idan source text ya canza, search index ma dole ya sabunta. Tsarin kamar Elasticsearch yana riƙe indexed data a segments kuma yana merge su a hankali; update ba yana nufin sake rubuta kowace posting list nan take a wurinta ba. Search visibility yana kuma dogara da refresh behavior na engine.

Saboda haka, lissafa ƙarin storage, indexing cost da yiwuwar jinkiri tsakanin rubuta document da samunsa a search. Inverted index yana da amfani musamman ga term-based text search; ba maye gurbin duk indexes na exact lookup, ranges ko transactional access ba ne.

1. Fara da ƙaramin collection

2. Yi amfani da lists don amsa query

3. Sanya indexing da search su dace

4. Adana isasshen bayani ga tambayar da ake yi

5. Fahimci kuɗin riƙe shi sabo

Tambayoyi masu alaƙa