Skip to main content

Search Features

Simple Search is the default screen when you open the Seqlense Doc app from the main menu. It features a search bar and displays the latest documents added to our database.

How does it work?

Simple Search looks for the keywords you entered in the document titles using a "LIKE" selector.

Tips for efficient search:
Choose simple keyword(s), e.g., "MiCa Crypto."

Advanced Search is accessible via the sidebar from any screen of the Seqlense Doc app. It includes a search bar and various filters.

How does it work?

Advanced Search looks for the keywords you entered:

    In the title of the document, AND
    In the body of the document (note: this refers to the extracted text from the document).

You can also refine your query using filters, such as regulator name, country, language, or the type of file being searched.

Tips for efficient search:
Choose simple keyword(s), e.g., "MiCa Crypto."

How does it work?

Deep Search takes a different approach. Instead of searching only the extracted text from documents, it uses embeddings to represent documents in the database. Your query is processed both in our vector database and the text database, and results are weighted to show you the most relevant documents.

Tips for efficient search:
Deep Search works well with natural language. While keywords are supported, it performs better with constructed questions, such as:
    "TFR implementation specifications for a CASP in the EBA guidelines"
    "What elements of the ESMA guidelines should be taken into account when producing an investor profile under MiFID II?".

Write down your request in english for improved results.

Deep Search is context-sensitive: If you ask a question about "MiFID," it will show you documents about "MiFID," not "MiFID II." Be mindful of this when formulating your queries and provide as much detail as possible.

Deep Search does not have access to document titles: Avoid asking about a specific document. Instead, search for the document first and then use the document question feature to interact with it.

How does it work?

AI Search builds on Deep Search, enhanced by a large language model (LLM). It allows you to ask thematic questions, retrieves the most relevant documents, and provides an answer based on the retrieved documents.

Tips for efficient search:
AI Search works well with natural language. While keywords questions ARE NOT supported. We advise you to use constructed sentences to yield better results, such as:
    "TFR implementation specifications for a CASP in the EBA guidelines"
    "What elements of the ESMA guidelines should be taken into account when producing an investor profile under MiFID II?"

DO : "What elements of the ESMA guidelines should be taken into account when producing an investor profile under MiFID II?"
DON'T : "ESMA MIFID GUIDELINES"

Try to be as precise as possible for best results while not reffering to specific documents as it does not have access to them.

Write down your request in english for improved results

AI Search is context-sensitive: If you ask a question about "MiFID," it will show you documents about "MiFID," not "MiFID II." Provide as much detail as possible for accuracy. Feel free to try it using the examples above.

AI Search has limitations:
        It does not handle versioning of documents or regulations.
        It is meant to help you explore a specific topic or answer questions.

AI Search does not have access to document titles: Avoid asking about specific documents. Instead, search for the document first and then use the document question feature to interact with it.

Back to top