Google Search now understands conversational language thanks to a fresh infusion of machine learning. Google’s Pandu Nayak elaborated in a recent blog post how a new development is one of the biggest leaps in the history of Google Search in the past five years.
The new update can now understand natural language. For instance, Search can now understand when you state qualifiers like ‘from’ and ‘to’, resulting in results that are much closer to what you’re actually look for. All this while, Search would basically ignore these terms and words and focus only on the keywords from your query.
This is achieved thanks to BERT, or Bidirectional Encoder Representation from Transformers, a technique for training neural networks that helps a system understand context in relation to the entire sentence or phrase. Previously, the system would only understand one word in relation to the one next to it.
Google Search is also being improved through better hardware. Google is deploying cloud TPUs to train neural networks handling search queries. Below is an example of what the change will mean for the user —
The change in Google Search is subtle and not very easy to see, but it goes a long way in providing meaningful search results. Google also said it will be using BERT to improve Search with other languages.
from Latest Technology News https://ift.tt/2piewd3
Comments
Post a Comment