We have improved the core of the intellectual search in AI-UniBot Personal Assistant & Corporate Chatbot by launching a multilingual semantic index powered by Artificial Intelligence!

This technology addresses a crucial problem in corporate knowledge management: the high cost and lengthy duration of indexing documents in a multilingual environment. At the same time, search accuracy remains uncompromised.

Previously, all documents were translated into English for a unified index, which required significant computational resources. Now, Administrators can choose one primary language and up to three fundamental (additional) languages, for which indexing is performed without translation. Other languages are automatically translated only into the primary language. As a result, the volume of operations is reduced by 70%–99%. During a search query, the system analyzes the context in all fundamental languages simultaneously, ensuring maximum relevance regardless of the original or query language.

We have processed the following key use case: Until a certain time, the International Holding used russian as its operating language. Later, the decision was made to fully transition to Ukrainian while simultaneously expanding into English-speaking markets. In this situation, the multilingual index became a crucial tool.

The Administrator set Ukrainian as the primary language and English and russian as fundamental languages. This eliminated the need to translate russian-language archival documents and newly created Ukrainian-language files. As a result, indexing costs were reduced by 75%, and the knowledge base update time dropped from 12 hours to just three. Meanwhile, the Employees receive accurate answers to queries in Ukrainian or English, even if the document was created in another language.

The technical implementation works as follows: When configuring, the AI-UniBot Administrator specifies the primary language (for example, Ukrainian) and the fundamental languages (such as English and Polish).
The system analyzes each document upon upload. If the file’s language matches one of the fundamental languages, the bot indexes the original. If not, it translates only into the primary language.

Thus, as soon as a User submits a search query, AI-UniBot determines its language and searches all fundamental indexes simultaneously using cross-lingual vector embeddings and keyword translations. Finally, the bot ranks the results considering semantic proximity to the query, regardless of the document’s original language. Limiting to one to three languages ensures a balance between speed and quality.

In short, the multilingual semantic index in AI-UniBot Personal Assistant & Corporate Chatbot dramatically reduces the cost and processing time of multilingual data. Search accuracy remains high, enabling Companies to efficiently work with both historical archives and modern documents and international sources within a single intelligent system.