Contents x
    3. Customize Hybrid Search
      • PDF
      Contents

      3. Customize Hybrid Search

        • PDF
        Article summary

        If you reach your limits with the feedback function, you can use Hybrid Search to generate more suitable results.

        Here you can find out what this means - and when you should use which setting for the AI.

        Search functions and also AIs distinguish between 3 settings:

        • Keyword search

        • Semantic Search

        • Hybrid Search

        Adjust the slider from left to right and observe what results are generated for your question.

        ❓ What happens during which search?

        What happens during a semantic search? - 1 (100%) on the slider

        • The search understands the meaning of your query – even if the words in the text are completely different.

        • The AI understands not only what you say, but what you mean.

        Example:

        You're searching for: "Laptop with long battery life"

        Texts will be found in your product data with phrases such as:

        • "lasts the whole working day"

        • "energy-efficient computer"

        • "Endurance device for on the go"

        🟢 Semantic search is good when:

        • You want to express yourself naturally.

        • The texts use many synonyms or colloquial formulations.

        • If you're looking for an idea or meaning, not concrete words.

        🔴 May be wrong if:

        • You are looking for something very precisely.

        • You only want to see certain terms (e.g. product names, technical features, legal texts).

        What happens during a keyword search? - 0 (0%) on the slider

        • The search will only look for exact words or very similar terms to the ones you have entered.

        • The search only understands what you say, not what you mean.

        Example:

        You're looking for: Laptop with long battery life

        Only texts in your product data with exactly these or very similar words will be found:

        • Laptop with long battery

        • Laptop with long battery cable

        🟢 Good if:

        • You know exactly what terms are in the text.

        • The texts are very technical or standardized.

        • You're searching for technical terms or product names (e.g., "x7-1365U").

        • You are looking for legal or formal texts with fixed wording.

        • If you know exactly how something is formulated.

        🔴 Not good if:

        • For example, the text is "enduring computer" instead of "laptop with long battery life".

        • Then it is not found because the words do not fit exactly.

        What happens during a hybrid search? - between 0.2 (20%) and 0.8 (80%) on the slider

        • You enter a question or term – the system searches for exact terms (keyword) and similarity of meaning (semantics) at the same time.

        • With the slider you control whether the focus is more on "exact word" or "meant meaning".

        Example:

        You are searching: "Laptop for on the go with touchscreen"

        • One document says: "Ultrabook with 1.1 kg weight, 12 hours runtime, touch-sensitive display"

        • Keyword search alone: Maybe you don't think so – "touchscreen" is not explicitly there.

        • Semantic search alone: Understands "on the go" and "touchscreen" – but may overlook the exact word "laptop".

        • Hybrid search (e.g., 60%): Combine both – you get exactly this entry because both match: concrete and analogous.

        🟢 Good if:

        • You are not sure how something is worded in the text.

        • You know a term, but you are also looking for the context.

        • You want a balanced mix (e.g. "lightweight laptop for travel").

        • You need both precise and flexible results.

        🔴 Not good if:

        • You only need very specific terms (e.g. "RTX 4070").

        • You're just looking for general meaning.

        💡 Note on Hybrid Search

        In order for you to get the most out of the hybrid slider, you or one of your work colleagues should know two things well:

        1. The type of product content

          1. Do you know how your products, features or offers are described in the PDFs or Excel files?

          2. Are they more technical terms, such as "Intel i7-1365U", or more general formulations, such as "very powerful"?

        2. Think about how your customers will search and ask now or later:

          1. Rather specific ("Does the laptop have an i7-1365U?")

          2. Or rather free ("Which laptop is fast and good for work and travel?")

        📌 Tip

        • If you are unsure, start with a value of 100% semantic search and test typical customer questions.

        • This will help you quickly find out whether you need more precise keywords or more semantic understanding.

        • To improve the search setting, you can rate the answers in the AI Evaluation Sheet (Excel).

        • Write down your searches and the expected answer and evaluate the AI answer with a score from 1 to 4 (1 = fits perfectly | 4 = doesn't match at all)

        👉 If at a certain point you don't see any progress in the quality of the results and are not satisfied yet, check out the next articles like Context Size to train your AI.

        📝 Examples of different searches

        I need a laptop for traveling – light and durable

        🔍 Keyword search

        ❌ Difficult: You search for "light" or "travel", but the text may say: "Ultrabook with 1.1 kg weight and 12h battery life". These terms are not literal – so you won't find anything or too little.

        🧠 Semantic search:

        ✅ Very good: The AI understands that you mean "light and enduring", even if it is worded differently.

        ✅ Best Solution:

        Semantic search or hybrid at 70-90%

        I want a laptop for gaming with RTX 4070 and at least 16 GB of RAM

        🔍 Keyword search

        ✅ Very good: You search for very specific, technical terms ("RTX 4070", "16 GB RAM") – which is exactly what you will find in the texts.

        🧠 Semantic search

        ❌ Not ideal: The AI may understand "gaming laptop", but in case of doubt ignores the exact model name.

        ✅ Best Solution:

        Keyword search or hybrid at 10-30% – high precision is important here.

        I need a device for graphic design – strong processor, good color representation, large screen

        🔍 Keyword search

        ❌ Partly inaccurate: You would have to know exactly whether it says "color space coverage", "sRGB", "retina" or similar. If you don't hit that exactly, you won't find anything.

        🧠 Semantic search

        ✅ Works well: Even if you just type in "good colors" or "design work," the AI understands the meaning behind it.

        ✅ Best Solution:

        Hybrid at approx. 40–60% – gives flexibility and still catches relevant terms.

        I want to know which laptop in my list is the most energy-efficient

        🔍 Keyword search

        ⚠️ Rather weak: If you enter "energy-efficient" but the text says "energy-saving" or "low consumption", it will not be recognized.

        🧠 Semantic search

        ✅ Helpful: The AI recognizes synonyms or indirect indications of energy consumption.

        ✅ Best Solution:

        Semantic or hybrid at 70-90%

        I'm looking for a laptop with a touchscreen that is also suitable for working on the train or café

        You are looking for a specific feature (touchscreen) – but also for a usage scenario (mobile working, on the go, being flexible).

        🔍 Keyword search

        ❌ Problem: Finds texts with "touchscreen", but ignores descriptions like: "for flexible working on the go" or "compact convertible for mobile professionals"

        🧠 Semantic search

        ⚠️ Not optimal: Understands the idea of "working on the train" or "mobile", but may miss models without an explicitly mentioned touchscreen because the meaning becomes too general.

        ✅ Optimal:

        Hybrid search hybrid at about 40-60%. Reliably takes the concrete term "touchscreen" with it.

        Was this article helpful?
        What's Next