media update’s Adam Wakefield decodes the relationship between contextual relevance, natural language, and AI, and why it is important for the advancement of the technology's ability to understand written text.

Understanding context starts by solving a language problem

Venkat Srinivasan, CEO of AI firm RAGE Frameworks, wrote in an article for the MIT Technology Review that context is a challenge because for a machine to understand it, it has to solve a language problem.

“Natural language text needs to be processed in the right context. The right context can only be developed if the technology focuses on the language structure, not just the words in the text, as most technologies seem to be doing,” Srinivasan says.

Machines should move beyond converting natural language in to data, which Srinivasan sees as natural language processing (NLP). Instead, they should focus on the linguistic structure of language and its principles. This is what he calls natural language understanding (NLU).

“In my view, NLP has come to symbolise the mechanical approach to natural language through conversion of text into data,” Srinivasan argues.

“Our real goal in AI is devising mechanisms for understanding the meaning of the written text.”

NLU requires reverse engineering text to its fundamental ideas to understand how those ideas were connected to form sentences and paragraphs. It's not possible to do so without understanding the context in which language is used, this would not be possible.

Understanding context, he says, has four aspects:
  • The ambiguity of language;
  • The use of  jargon, be it news reporting or legal writing, within a given context;
  • The use of synonyms to mean the same thing; and
  • A document might refer to something that is not explicitly included in the text;

To understand context, machines will draw on existing knowledge

The key to AI meeting these challenges is, in Srinivasan’s view, to create a repository of global data that can be “retrieved, in context, to supplement the text in the document to gain full understanding of the meaning of the text”.

In other words, AI needs to leverage Big Data and draw insights from it to understand the contextual relevance or the context in which language is used.

Understanding context is having prior knowledge on a given topic to decode what is being said at a specific time. In a conversation with a friend, the other party has all previous interactions to draw on to weigh, measure, and decode what their friend means and says.

The different technologies that make up AI are becoming better at storing and sifting through vast amounts of data. This will enable machines to have previous knowledge to draw on to understand the context of used language. This, combined with breaking down language to its core elements through NLU, is allowing AI to go beyond just processing language but truly understanding its meaning.

Want to stay up to date with the latest media news? Subscribe to our newsletter.
Classification is an important part of how AI learns the difference between objects and concepts, #ArtificialIntelligence: What is classification?