x

Search the Findwise web

All you need to know about Machine Learning

Transform your documents into an asset – instead of a burden

How to use AI to automate translation and perform metadata extraction on documents

Challenge

In the changing landscape of IT, companies invest more and more in new communication channels like social media or chatbots, but still the majority of” company knowledge” is stored in documents.

For organizations working on many markets, with different languages, the translation issues just ads complexity. Internally within an organization it might be a solution to select a “corporate language” and keep all documents in that language. But if we are talking about specific types of documents and information that are supposed to be used by customers or citizen in a specific country it is rarely accepted to use anything else but the native language.

Think about all the marketing material, installation manuals, spare part specifications, instructions, policies etc. All of these documents are most often published and distributed in the native language – creating a vast cost and effort from the publishing organization to translate, but even more problematic – to store and keep track of.

Solution

We have used Artificial Intelligence (AI) to automate the process of translation and to perform automatic metadata extraction on documents.
(The picture below shows an overview of the solution)

Using AI in the solution has two major benefits:

Firstly, the process of translation itself can be greatly improved. But it also makes it possible to add corporate “terms and phrases”.” Thus, it is possible to reuse translations and make sure that the right naming is used – throughout the organization.

Secondly, the adding of metadata to each document creates the foundation of a good search solution, making it possible to actually “find what you are looking for”. This can of course also be done manually by users – and it has been tried many, many times before - with limited results. When AI can identify and add correct metadata to a unique document the result from a user and search perspective is tremendous.

Some of the features included in the solution are:

  • Automatic translation of a document to English or any language selected
  • Ability to select and translate one, or many documents, to multiple languages
  • Custom translation dictionary for hermetic and specific names
  • Metadata extraction based on document structure and content – we separate structure and content making it possible to keep the structure of a document
  • Metadata extraction model (or models) – which makes it possible to customize and select which metadata that are extracted
  • Search interface with filters
  • A Teams chatbot understanding user´s natural language making it possible to ask the bot (another way of searching) for the right document in your own, local, language

The solution is built on Microsoft Office 365/SharePoint Online and Microsoft Azure platform using many different “products” like SharePoint Online, Microsoft Flow, Azure Cognitive Services, Azure Bot Framework and Microsoft Teams. The solution is customizable and accessible on various devices (i.e. phones, tablets, PCs).

Result

A solution like this can have a positive impact on many different levels and processes within an organisation. Our focus has been to create the following business values:

  • Time and cost savings: Automatization of every day processes
  • Higher quality of services: Reuse of translations and common language
  • Information as an asset: people find what they are looking for, in their natural language

Technologies used on Transform your documents into an asset – instead of a burden

We build search solutions based on any platform and have partnerships with all leading vendors.

Microsoft SharePoint & Office 365 - Search in SharePoint and Office 365 is more powerful than ever before