Documents and knowledge bases
Create a knowledge base, upload literature and set its scope.
- Last updated
Create a knowledge base
Open Resources → Knowledge Assets and select New knowledge base. Enter a name and description, choose a type and visibility, then save.
New bases currently support Perovskite. Other and Wide-bandgap types are not yet available. Use a name that describes the collection, such as “Perovskite interface passivation papers,” so it is easy to select in a conversation.
The creation form offers two visibility settings:
- Private: the knowledge base owner can view it.
- Research group: active members of the selected group can view it.
Sharing scope: the interface says perovskite extraction results are synchronized to the device database. The current creation and upload forms offer no switch to disable this. Separate access rules for original files, passages and in-base parameters, and visibility of synchronized device data, remain unspecified. The Private setting therefore does not establish that synchronized data is private. Wait for these rules to be clarified before uploading unpublished material.
Upload documents
Open Documents in the knowledge base and select Upload documents. Choose files or drag them into the upload area. Supported formats include PDF, Word, PowerPoint, Excel and text. Upload 1–20 files at a time, each nonempty and no larger than 100 MB.
Check the status of each file. If only some uploads fail, retry those files instead of uploading the successful ones again.
A completed upload still needs parsing. Transfer completion means the file was received. Parsing, indexing, parameter extraction and graph construction may still be running.
Inspect extracted content
Check the processing stages in the document list and refresh when needed. Once processing finishes, open the document to compare its summary, source passages and extracted parameters with the original.
Chinese interface. Check each stage separately. This public-paper example has completed processing and produced 20 source passages.
PSPP means Processing, Structure, Properties and Performance. Follow these relationships in the graph to find supporting passages.
See Parsing, parameters and evidence for processing stages, and PSPP knowledge graphs for paper filters and relationships.
If parameters or graphs are empty, check processing status. Once processing ends, use the available retry action if needed.
Extraction can miss units, conditions or context. Check the corresponding source passage or paper before using a parameter.
Open Evidence text in the document details to read source passages associated with the extraction.
Chinese interface. Check values together with methods and test conditions. Source: Xiong et al., licensed under CC BY 4.0. The image shows text parsed by the platform.
Use a knowledge base in Research
Open Research, select Knowledge bases below the message box, choose the relevant base and ask a question:
Select up to 10 knowledge bases per message. If retrieval finds nothing, check your selections and document indexing status before changing keywords.
Follow Retrieve literature evidence in a conversation for a complete question, source verification and follow-up exercise.
Delete documents or a knowledge base
The deletion dialog says that the document and its associated parameters and chunks will be removed irreversibly. The interface does not state whether synchronized device records are also removed. A knowledge base must contain no documents before it can be deleted.
Keep a copy of an original file if you need it. To exclude a base from one query, deselect it in the conversation; you do not need to delete its documents.