Features Enhancement for Ecommerce Search

Allow customers to find exactly what they want with improved search results.

The need for structured data extends to all kinds of search, be it standard free-text search or new-generation voice search. In ecommerce, that implies every laptop in your catalog needs to have its screen-size expressed in inches and every shirt needs the right color tag. Without relevant information it is difficult for consumers to find exactly what they need.

Our Feature Enhancement API fulfills this need using a combination of methodically curated taxonomies, artificial intelligence techniques and strategically inserted heuristics.

Key standardizaion

Key Standardization to maps all attribute names to our standardized taxonomy.

Display size: 13 inches

{  "screensize": "13in"  }

Approximate size 2 inches x 3 inches x 4 inches

  "length": "2in",
  "width": "3in",
  "height": "4in"

Average Battery Life: 8 hours

{  "batterylife": "8 hours"}

Named Entity Recognition

Named Entity Recognition to extract structured key-value pairs from unstructured fields such as product title and description.

Description: This laptop has an Intel Core i7 processor

{  "processortype": "Intel Core i7"  }

Name: Minnie Mouse Toy by Disney

{  "character": "Minnie Mouse"  }

Description: Painting by Van Gogh from the year 1831

  "artist": "Van Gogh"
  "year": 1831

Value Normalization

Value Normalization to map attribute values of selective fields to comparable formats.

length: 2in, width: 3in, height: 4in

  "length": "50.8mm",
  "width": "76.2mm",
  "height": "101.6mm"

weight: 1lbs

{  "weight": "453592mg"  }

operatingsystem: mac osx v10.9

{  "operatingsystem": "Mac OS X v10.9"  }

How to get setup

Tell us the categories and features/attributes you’re interested in. Next, send us the dataset you’d like standardized and normalized. If you don’t have your own dataset, tell us what kind of data you’re looking for.

After a brief consultation, our technical leads map your category-attribute list to our internal taxonomy. Then, start testing our algorithms with a Pilot or Proof-of-Concept.

Semantics3 products in this solution