Auto Classification of Documents in SAP Enterprise Portal

[ZJamal]

Hi experts,

A new index has been created in SAP Enterprise Portal and assigned a repository containing small set of folders along with documents. The index has been configured for auto classification of documents. Initial classification of documents was performed manually to train the categories. Afterwards, when a new document is uploaded in the same repository (which has been used for newly created index), the system should auto classify document that is similar to the example documents into the categories of the taxonomy. When a new document is uploaded in KM in same repository used for the said index, crawler service does not classify this new document in any category of the taxonomy. However, the newly uploaded document is available in Classification Inbox under Documents to Classify folder (can be manually classified).

Newly uploaded documents are not classified by the system even by using Auto Classify option under classification inbox (Content Management). However, manual classification works.

I would highly appreciate if anyone could guide how to enable SAP EP system for auto classification of newly uploaded documents in KM.

Regards,
Zahid Jamal

[ZJamal]

Hi all,

We are stuck with this issue, please guide.

Anyone tried auto classification of KM documents in EP?

Regards,

Zahid

[cathal_kelly]

Hi Muhammed,

By your reference to manually training the documents, I assume you are using example based classification in this scenario rather than Query Based classification? In this case, documents will only be automatically classified if the collection of training documents are of a sufficient quality and scope to determine the classification of the newly uploaded documents. Are you encountering a scenario whereby no documents are being classified (even those which are almost identical to one or more of the training documents)? Or having an issue whereby some of the documents are not being classified but other are being classified successfully?

If it's the latter, then I would suspect that there are perhaps not enough documents of sufficient similarity within the collection of training documents.

If no documents are being classified then I would question whether the indexing is running. 'Auto Classification' still requires the indexing to run on the documents in question and they are then classified as part of this indexing process. Is the IndexServiceTaskQueueReader assigned to a valid CM system and showing as being in status 'running' in the component monitor (System Admin > Monitoring > KM > Component Monitor)? This service can be found under the services > scheduler entry in the monitor. Are the documents being indexed successfully?

Kind regards,

Cathal

[ZJamal]

Hi Cathal,

Thanks for kind feedback.

We are trying example based classification. The issue is only with automatic classification which some how system is not performing either due to insufficient number of training documents or due to quality of training documents. I could not find any clear guideline regarding total number of training documents required for successful auto classification of documents by the system.

Indexing is working fine. IndexServiceTaskQueueReader service is assigned to a valid and only one CM system. We successfully applied classification by manual mapping of document to their categories.

Please guide.

Thanks and regards,

Zahid