Configuration
OfficeDocumentToTextAndImages Filter Plugin

Introduction

The FilterPlugin.OfficeDocumentToTextAndImages is used to filter Office documents, convert them to PDF documents to be available for the PDF preview and generates images for every page in the converted PDF document.

The generated page images can then be retrieved with the answers from the similarity search (if configured).

This is how the FilterPlugin.OfficeDocumentToTextAndImages plugin processes the documents:

1. The document is dispatched to the FilterPlugin.OfficeDocumentToPDFContentFilter to
   1. convert to PDF,
   2. dispatch the PDF to the FilterPlugin.PDFPreviewFPDFFilter to
      1. generate the PDF preview,
      2. generate the thumbnail and
      3. generate images for every page
2. Next, the document is dispatched to the Content Filter Plugin with the highest priority to
   1. generate the searchable text content.

For more details, see section “Configuration”.

The filter plugin can be applied to the following file extensions:

Configuration

This filter plugin is deactivated by default or must be explicitly activated.

Base configuration

To enable this plugin, go to the Configuration menu in the Management Center and open the Filter tab. Expand the settings of the filter service which should use the FilterPlugin.OfficeDocumentToTextAndImages. Scroll down to Content Filter Services and add the following plugins:

• FilterPlugin.OfficeDocumentToTextAndImages
• Any other plugin which should filter your office documents and produce the text content, e.g. FilterPlugin.Pandoc or FilterPlugin.ApacheTika

Set the priority of your Content Filter Plugins:

• FilterPlugin.OfficeDocumentToTextAndImages: Should have the highest priority. Alternatively, you can manually select the FilterPlugin.OfficeDocumentToTextAndImages for each extension where this should be used.
• The other filter plugin used for filtering the content must have a lower priority
• 

Example use-case: retrieving slides of a Microsoft Powerpoint presentation as images

To be able to retrieve the page images from the similarity search (NLQA) along with the answer coming from this page, some additional configuration steps are required. For more information, please refer to Configuration - Filter Plugins and Documentation - Mindbreeze InSpire.

For example, if you want to extract the slide images in addition to answers from a Microsoft Powerpoint presentation, you need to configure the filter and index as follows:

1. Configure NLQA
2. In the Filter settings, configure:
   1. FilterPlugin.OfficeDocumentToTextAndImages as highest priority for .pptx documents
   2. FilterPlugin.Pandoc as second highest priority for .pptx documents
   3. In the FilterPlugin.Pandoc configuration, configure:
      

1. In the Index settings, configure:

1. In the Client Service settings, configure:

To verify if the configuration is correct, open the developer tools in your browser and trigger a similarity search in your Insight App. The response should contain:

answers[].text.pages[].page.image

e.g. with value:

"ref_id": "tag:mindbreeze.com,2026:value-ref/key-to-your-file:2:mes:page_image_13

Also see api.v2.search Interface Description - answers.text.pages.

Troubleshooting

FilterPlugin.OfficeDocumentToTextAndImages is not used

If the FilterPlugin.OfficeDocumentToTextAndImages is not for filtering your documents as intended, it is most likely the priority settings of your content filters are incorrect.

Check app.telemetry

You can verify that in the app.telemetry Filter Service logpool. Select the filter request for the document you want to troubleshoot and open the Associated Requests. E.g. if you want the FilterPlugin.OfficeDocumentToTextAndImages to be used together with FilterPlugin.Pandoc for a PPTX documents, the Requests from Content Filter Service should look like in the screenshot below. The FilterPlugin.OfficeDocumentToTextAndImages should be the first content filter to be called, then the document is dispatched to the other content filters in the filter chain.

Check Filter Service plugin-mapping.txt

Another way to verify the effective Content Filter priorities is the plugin-mapping.txt file in the Filter Service logs (/data/logs/log-mesfilterservice-23400/current/plugin-mapping.txt), see example below.

Extension: pptx

GroupID: FilterPlugin.OfficeDocumentToTextAndImages Priority: 9223372036854775807

        id: FilterPlugin.OfficeDocumentToTextAndImages extensionPoint: mindbreeze.filter.ContentFilterService priority: 20000 content_crc_calculation  registered

GroupID: FilterPlugin.Pandoc Priority: 20001

        id: FilterPlugin.Pandoc extensionPoint: mindbreeze.filter.ContentFilterService priority: 20001 registered

GroupID: FilterPlugin.ApacheTikaWithThumbnails-Latest Priority: 11100

        id: FilterPlugin.ApacheTikaWithThumbnails-Latest extensionPoint: mindbreeze.filter.ContentFilterService priority: 11100 registered

GroupID: FilterPlugin.ApacheTika-Latest Priority: 10050

        id: FilterPlugin.ApacheTika-Latest extensionPoint: mindbreeze.filter.ContentFilterService priority: 10050 registered

GroupID: FilterPlugin.OfficeDocumentToPDFContentFilter Priority: 10000

        id: FilterPlugin.OfficeDocumentToPDFContentFilter_1 extensionPoint: mindbreeze.filter.ContentFilterService priority: 10000 content_crc_calculation  registered

GroupID: FilterPlugin.MetadataOnly Priority: 50

        id: FilterPlugin.MetadataOnly extensionPoint: mindbreeze.filter.ContentFilterService priority: 50 registered