How the bot searches for the answer¶
Discover how glacom.AI finds the right information among your documents to answer the customer. This helps you understand why the bot chooses certain answers and not others.
How it works¶
The system doesn't just search for exact words, it understands their meaning. If a customer asks "What time do you close on Friday?", the bot doesn't just look for the word "close", it understands that the user wants to know the closing time.
glacom® uses hybrid search: 1. Semantic search: matches the intent of the question with the concepts present in your documents. 2. Keyword search: essential for finding product codes, proper names or precise technical terms.
The bot doesn't read the entire archive every time. It selects only the chunks most relevant to the question and uses them to build the answer. This process happens after the indexing of your content.
What you need to do¶
Nothing: the search process is automatic. To improve results, make sure to write content clearly.
What glacom® does¶
The system transforms the customer's question into a mathematical concept, compares this concept with your data and retrieves the closest text fragments. Finally, it passes this information to the generation phase of the answer.
Limits¶
If the information is written too ambiguously or is contradictory across multiple sources, the bot might struggle to choose the most correct fragment. If it doesn't find any block of text close enough to the meaning of the question, the bot will answer that I don't know.
Frequent problems¶
- The bot can't find a specific product code? → Check that the code is written exactly the same way in your documents.
- The bot confuses two similar services? → Distinguish the two services better in the texts, using different names and specific descriptions.
- The answer is too generic? → Add more details to your content to give the bot more "blocks" of precise information to retrieve.
From the app
Not available from the app: the search logic is internal to the glacom.AI platform.