New
- Automatic renaming of source files in export
- Renaming documents in plusmeta is now possible fully automatically.
- A metadata is configured for this. This can be composed of other metadata and also inherit metadata from digital twins that are linked to the document.
- The source file is renamed in the export according to the value of this metadata.
- The function also works for documents that are linked to several digital twins. If VDI 2770 packages are generated according to the digital twins, the renaming metadata is specifically updated in the export for each digital twin and the associated source file is renamed.
- Please contact your project manager if you would like to use this function.
- Extraction of technical data with variable units of measurement
- There are several possible units of measurement for many technical data. Thanks to the new function, plusmeta can now recognise this technical data with variable units of measurement and extract it as metadata.
- Several units of measurement can now be stored in plusmeta for a physical quantity. This physical quantity is then assigned to the metadata.
- For example, values in ‘mm’, ‘cm’ or ‘m’ can be found for the metadatum / physical quantity ‘length’.
- This means that the automatic recognition of such technical data is now possible, particularly via dynamic extractors and auto-extractors.
- Mass export of documents from projects
- It is now possible to download all objects (documents) of one or more projects in one action.
- A directory structure can be specified on the basis of metadata. The documents from the projects are sorted into the directory structure according to the metadata.
- The source files of the objects can also be renamed in the export from the project view using a configured metadata.
- Optimised rule-based recognition
- Foreign-language labels can now be analysed differently in rule-based recognition.
- This is to prevent metadata with similar multilingual labels from receiving a disproportionate number of points in metadata recognition.
- Configuration options in plusmeta Insights
- With plusmeta Insights you can chat with your project documents and metadata.
- A new configuration object for Insights enables versatile configuration options for the function.
- Among other things, the configuration object can be used to set which LLM is to be used (e.g. Claude, Chat GPT, Llama 3 etc.).
- In addition, the rule-based metadata recognition on the text of the prompt can be fine-tuned. This ensures that keywords and extracts from the plusmeta Knowledge Graph are made available to the LLM in order to provide more precise answers.
- As a further example, it is possible to configure the extent to which the chunks within the prompt should be given more context.
- Please contact your project manager if you would like to use this function.
- Optimised automatic linking of objects
- The logic behind the automatic linking of objects is now configurable.
- Regular expressions can be used to specify when two documents with a specific prefix or suffix should be linked.
- Automatic linking is only carried out if the project setting is activated.
- Optimisations of the OCR function
- Character recognition now indicates the estimated processing time for each document.
- It is now configurable on which pages or page areas the character recognition is applied. This means that drawing headers, for example, can also be specifically read with OCR.
Other
- Segmentation can now be configured so that segment boundaries only occur at page boundaries.
- Write-protected PDFs are now labelled in the plusmeta interface
- The LLM ‘Llama 3’ has been added to plusmeta.
- plusmeta now supports more date formats.
- Original files from the PDF/A conversion no longer end up in a directory in the VDI 2770 export, but are now given the prefix ‘original_’.
- There are new export functions in the similarity analysis, such as the filtered export and the JSON format.
- The recognition of PDF/A conformity has been improved.
Bug
- When the page is reloaded in the browser, the character recognition is no longer restarted completely.
- Microsoft Help Viewer 2.0 packages are generated correctly again.
- A performance problem caused by Chinese characters has been fixed.
- Some errors in plusmeta Insights have been fixed.
- An error in LLM data extraction has been fixed where the response was incorrectly output as a JSON structure.
- A display error with objects of the type ‘object relationship’ has been fixed.
- An error in the generation of VDI 2770 packages with digital twins has been fixed.
- The presumed PDF title is now also determined correctly for OCR documents.
- An error in CSV export with multi-value metadata and files with spaces in the name has been fixed.
