Context Retriever
Context Retriever
Context Retriever matches the users’ question with information contained in documents and retrieves different chunks or sections from across one or more documents. This process will allow business users to verify if the right user documents are being retrieved.
It delivers more specific information and allows further filtering of search results. It is a search methodology that returns portions of the document that match the query. The search results (chunks) are ranked and scored based on the most relevant answer to the query.
 |
Click Search and select Context Retriever.
Enter the Question. For more specific information, the advanced search allows you to further refine the search results using the predefined fields.
Advanced Search:
Click the Filter icon to view the Advance Search fields.
The "Select or add an item" dropdown displays a list of metadata fields, which can be user-defined or automatically generated by the AI during document enrichment.
The "Select operator" dropdown shows characters that help refine search queries. Operators enable the search engine to retrieve documents where the metadata field value matches the specified value.
Enter the value to obtain more specific and relevant results.
Click Apply Filter.
Note: Users can add multiple filters or conditions to further narrow down the search results.
Select the type of Retrieval from the dropdown.
Semantic Search compares your query to stored documents by measuring how close they are to each other. It then shows you the top matching results.
Semantic Search + Reranker utilizes the same retrieval process as Semantic Search, but it also rechecks the top results to make sure the most relevant ones appear at the top of the list.
Enter the K Value. K Value refers to the number of chunks (search results) to display.
Click the lens icon.
The chunks (search results) are displayed according to their rank and score, which are based on the relevance of the answer. The score determines the ranking, with the chunk having the highest score appearing first. Additionally, the name of the source document from which each chunk is extracted is displayed at the bottom of every chunk.
 |
Search Results Summary (Context Summary)
The search results summary enables you to view the Rank, Document Identifier, and Document Titles from which chunks are retrieved. Also, it enables you to provide feedback on the summary using the thumbs up or thumbs down icons.
 |
Summary Feedback
You can provide feedback on the summary If you disagree with the summary, click the thumbs down icon. A pop window appears. Select a reason whether it's due to 'Incorrect documents' or 'Incorrect ranking'. Select the correct documents from the drop down.
If you agree with the summary, click the thumbs up icon. A pop window appears. Select a reason from the drop down.
The System stores the elements from user feedback, which will be leveraged to improve future search accuracy.