StatusNeo

Semantic Search? Oh, I was already using it!

What is Semantic Search?

• Semantic search is a searching method or technique.
• In this technique, not only the searched term but the texts or words having similar meanings are also searched.
• As the word semantic suggests- relating to meaning in language or logic. (by oxford dictionary), it is clear that the context is searched not only the searched term.
• According to Wikipedia Semantic means-  the study of reference, meaning, or truth.
• All these definitions of semantic explain what semantic search really means.
• In short, semantic search is the search for the contextual meaning of the searched term or query.

• In the above example, we can see that when we are searching for what to read, google search is suggesting to us the name of books.
• This is an example of the semantic search, on what to read was taken by its contextual meaning, rather than just the keywords.

• Similarly in the above image, we can see the results of what to watch this weekend, is showing us the list of movies and shows.
• I hope by now it must be clear what semantic search is all about.

Are we already using it, in our day-to-day life?

• By now you must be wondering where else you have been using semantic search in your life, so let’s make a few real-life examples from your mind.
• The most common example is online shopping.
• Online shopping brands such as Amazon, Flipkart, Myntra are using Semantic Search for their apps.
• Not only on the searches it is also getting used in the recommendation systems of the online shopping apps.
• Google search or internet search is also an easy-to-understand example of semantic search.
• Social media such as Facebook and Instagram are using it up to a great extent by showing us all the possible meanings of what we have searched for.
• In online shopping, say you want to buy a shirt, so you searched for it, then whenever you click on a shirt it shows you matching watches, jeans, and all.
• This is how the semantic search is mapped here, i.e. if someone is searching for clothing he might be interested in the accessories related to it.

• From the above example, we can see when we are searching for jeans, it is showing us results for all types of jeans irrespective of gender.
• But when you log in and you are going to search for the same, the result will vary according to your previous searches.
• We can conclude that there is a mapping between our search history to the query we are searching.
• Not only the search history but also our location, preferences, and contextual meaning is mapped to the semantic search.

This is how Semantic Search Works?

• I will share the basics of the semantic search algorithm.
• The main keys that are used in a semantic search are-
  • Text matching( synonyms, words with similar meaning)
  • User location(It plays a key role while showing the result)
  • Global and User Search History(It helps in understanding the preferences of the users)
  • NLP (Natural Language Processing Algorithms)

• Text matching is the first thing that is used, via the lexical matching engine.
• Based on the user location and history the ranked web pages are shown, these rankings keep on changing.
• Web-based semantic search works differently from a product-based search such as Amazon or Flipkart.
• Google uses Knowledge graph techniques to meet the best results for the web-based search.
• Google’s Knowledge graph is the Google database that has the dataset about people, places, and things.
• The knowledge graph was a milestone in the web semantic search history.
• It revolutionaries the search results.
• You can read more about it from – Google Knowledge Graph

• According to Google, this information is received from various trusted resources, such as Wikipedia.
• Now coming to the product-based Semantic search, let’s see how Amazon does this.
• Amazon has already published a paper on how they use semantic search for their products. Semantic Product Search
• At the basic level, amazon uses the Neural Network Architecture, Loss Function, Tokenization Methods to improve the results.
• In the below diagram, we can see Amazon’s System architecture for augmenting product matching using semantic matching

• From the below diagram we can see the Illustration of neural network architecture used for semantic search in Amazon.

Conclusion

• We can conclude that directly or indirectly we are always using semantic search in our life, either for shopping or for web surfing.
• I hope that this blog will help you to understand the basics of Semantic Search.
• In the upcoming blog, we will try to implement it.

Thanks for Reading!

References-

• https://assets.amazon.science/95/2f/51e4ebdb4347990fedc1815d7327/semantic-product-search.pdf
• https://support.google.com/knowledgepanel/answer/9787176?hl=en

Still Curious? Visit my website to know more!

For more interesting Blogs Visit- Utkarsh Shukla Author