Diving into sentence-level searching using neural networks at KMWorld Connect 2020

At KMWorld Connect 2020, Paul Nelson, innovation lead, search & content analytics, Accenture, presented proven and innovative solutions that perform search at the sentence-level, identifying the exact sentence within the document that provides the required information in a session titled, "Achieving Sentence Level Searching Using Neural Networks."

KMWorld Connect, November 16-19, and its co-located events, covers future-focused strategies, technologies, and tools to help organizations transform for positive outcomes.

According to Nelson, neural networks technology is changing the game for enterprise search and KM systems. Most search and knowledge management systems retrieve and manage documents, but for many use cases such as policies, procedures, testing guidelines, and standards documentation, document-level analysis is simply too coarse-grained to provide the user with the answers they are seeking. However, neural networks can achieve sentence-level search, even when the words specified by the user are not exactly the same as the words used by the author.

Nelson was an early pioneer in the field of text retrieval and has worked on search engines for more than 30 years. He was the architect and inventor of RetrievalWare, a natural-language based statistical text search engine which he started in 1989 and is now owned by Microsoft. Nelson also served as the chief architect at Search Technologies until its acquisition by Accenture in July 2017, and he is now in Accenture’s Applied Intelligence group.

According to Nelson, neural networks technology has flourished in the last couple of years and is getting bigger and bigger.

Through explanation and demonstration, Nelson showcased the use of neural networks technology in action for frequently asked question-answering and for sentence-level search, showing how it can bridge the gap between the user and content.

In using AI for KM, Nelson emphasized, it is important to remember that:

AI is more than just classification and entity extraction

• Now we are getting neural networks that can really understand text and we can use that for sentence level searching, connect content, or even summarize and rewrite content

Successful projects will include targeted AI

• Try to solve the problems you can solve today with AI, he advised. With sentence level searches, some are great, some are not. Don't try to expect perfection or magic and make sure there is a human in the loop.
• At the same time, you should gather the data you need to solve the problems of the future.

You still need a great search engine

• The results of AI feed and enhance the search engine.
• Search engines are the only way to scale.

Replays of KMWorld Connect webinars will be made available for on-demand viewing.