More than a thousand DocuWare customers are using Intelligent Indexing today for ultra-quick document storage. This unique service reads through documents and recommends specific indexing terms. Nearly 14 million documents have already been stored and indexed with Intelligent Indexing. The service’s accuracy rate is above 80 percent. But what can you do when Intelligent Indexing is not reading the words you want it to? It’s just a matter of the right training.
Here’s an example: You receive the first invoice from your new beverage vendor, Fresh Water Beverages Inc.. You scan the invoice and Intelligent Indexing takes care of the indexing. All of the database fields – record number, date, amount, etc. – are perfectly filled out. But you notice the name of the company is not complete: instead of the entire company name of "Fresh Water Beverages Inc.," only "Beverages Inc." appears in the index field.
Better to Click than Type
Intelligent Indexing is constantly evolving and learns through your corrections. Here’s the best way to help it learn: when entries are coming through incorrectly or incompletely, don’t use the keyboard to make corrections in Intelligent Indexing. Instead use the “One Click Indexing” feature. Otherwise the service won’t learn to find the right data from the document and will instead rely on manual entry for subsequent documents.
Here’s how it works:
1. Click on the affected index field – the field as well as the source for the index entry in the document will both be highlighted.
2. Now press on the mouse key and mark the complete text in the document – in our example, the entire entry of "Fresh Water Beverages Inc." The company name will immediately be transferred into the storage dialog.
3. You can now store the document. Intelligent Indexing will remember the correction you made. As a general rule, the next time you store an invoice from this beverage company, the indexing will be completely correct thanks to this tremendously clever service.
Please note: If for some reason a document was indexed and stored incorrectly with Intelligent Indexing, you may have to correct the entry a few times in order for the service to unlearn its previous choice and work correctly going forward.
Here’s a video about how to best train Intelligent Indexing:
Good Resolution – Great Results
The accuracy of Intelligent Indexing largely depends on a document‘s legibility. We therefore recommend scanning a document with a minimum resolution of 300 dpi. Otherwise it’s easy for so-called OCR mistakes to occur. Common examples: the system can‘t differentiate between "i" or "!" or "I"; or it turns a "w" into two "v".
Again, it’s best to correct these reading errors with One Click Indexing. Manual entries should really be the last resort – only to be used if you cannot capture the right information from the document. If you are noticing several OCR errors, then you might want to first adjust the resolution in your scan settings.
Additional details about this topic can also be found in these FAQ articles: What do I have to keep in mind when setting up the Intelligent Indexing Service? and How can I assist the Intelligent Indexing Service with learning index data?