Document-searching reinvents Knowledge Management

How do you make the most of a company's existing content, ranging from text to audio and video? How can you easily explore this content to build new knowledge? Demo of a totally new solution for semantic document management.

Offer a more effective search in all content formats including often-underused videos and audios.

In most companies, data is all over the place, often organised in silos, by entity, or by business segment. Finding any particular information can sometimes be an obstacle course in itself and can involve investigating numerous channels and sources, having to identify yourself over and over again, spending time watching or listening to an entire conference or reading the entire transcript.

Faced with this reality, Orange’s Data Search and Knowledge researchers, who have been working for several years exploring and developing multimedia Knowledge content, have produced a prototype of a multimedia document database for businesses. This unique and innovative solution called Sema4KM, which is a database-cum-search engine, uses speech recognition technology, semantic analysis and keyword extraction tools, as well as speaker analysis and indexation tools developed by Orange Research and combined to create value. The goal? A multimedia document management database that offers user-friendly navigation and a more efficient search engine for all content formats including often-underused video and audio. The operating engine also offers the ability to cross-link knowledge sources.

“We have invented the term ‘Explorescence’, to describe this new way of using content to create Knowledge,” says Géraldine Damnati, Automated Language Processing Research Engineer at Orange.

Featured video and audio content

The demo at the 2019 Salon de la Recherche is based on documents produced by Orange Research, a corpus composed of webinars, filmed conferences, blogs, internet & intranet sites, a Yearbook, and a whole raft of presentations… Based on requests for information about topics such as Blockchain, a search can quickly lead to a wide selection of additional content in various formats. In addition to these texts, new features include extracts from conferences dealing with broader topics, and the canvassing of consensus on those topics.

This is one of Orange’s new knowledge management tools, and its Data and Knowledge researchers are working hand-in-hand with the Knowledge Management and the Digital Business teams to tailor solutions to handle the content and other problems that businesses are facing. “We’re already thinking about developing the tool further. For example, searching by word or phrase similarity, multilingual search, and conversational searches that allow the user to ask questions to get even more precise answers.”

Read also on Hello Future

Two people are seated in front of a computer, discussing a project. Spools of thread are visible on the table.

How to avoid replicating bias and human error in LLMs

Discover

NORIA: Network anomaly detection using knowledge graphs

Discover

Machine learning shows fresh potential in urban renewal

Discover
GettyImages - machine learning - research recherche

Machine-Learning-Based Early Decision-Making (ML-EDM)

Discover

Machine learning to combat ocean plastic pollution

Discover

Green and Local Energy Exchanges and Optimizing Network Consumption

Discover
A man having his face scanned by a tablet for better audiovisual recognition

Better Hearing Through Seeing: Separating Voices in an Audiovisual Stream

Discover

Encrypting or classifying Internet traffic: do we have to choose?

Discover