Booking with Natural Language Processing (NLP): Understanding Customer Sentences

Classic bots work via menu: "1 — price, 2 — booking". Modern bots talk like humans. The difference in between is NLP — Natural Language Processing.

Intent and Entity

Intent: What the customer wants to do. Such as "booking", "cancellation", "price inquiry", "availability".

Entity: Specific pieces of information in the sentence. Date, time, number of people, activity name, location...

Challenges of Turkish NLP

• Agglutinative language: "rezervasyonum" (my booking), "rezervasyondan" (from the booking) — root finding is critical.
• Free word order: "Yarın 4 kişiyiz" (We are 4 tomorrow) = "4 kişi yarın geliriz" (4 people we will come tomorrow) — word order can change.
• Regional expressions: "öbürsü gün" (day after tomorrow), "haftaya pazartesi" (next Monday) — local calendar logic.

Context Management

The message "Okay, but is 10 o'clock available?" is meaningless on its own. The bot must remember the previous message (which tour, which date). A bot without context management interprets every message from scratch — bad experience.

Ambiguity Management

When a customer says "tomorrow", do they mean the following day or "by tomorrow"? If the bot isn't sure, it should ask a clarifying question. Proceeding with a misunderstanding is worse than an error.

Sıkça Sorulan Sorular

Does the bot understand every sentence?

No, not 100% but 94%+ accuracy. When it doesn't understand, it transfers to a human operator.

Can the WhatsApp bot answer in English?

Yes. WhatsApp message language detection and responses are limited to Turkish and English.