Data Science Solution for the Automotive Industry
By Amit Nahata
How do you find out more about new and used cars with all specifications, reviews and price details? Edmunds.com is an online portal for automotive information. It lets users research about new and the used cars. It offers full technical specifications, user reviews, vehicle comparisons and buying-advice from experts. Headquartered in Santa Monica, the company is almost half a century old. Shopping being the main business focus, it is essential for Edmunds to offer the users with accurate and relevant information on vehicles. Edmunds has to rely on Original Equipment Manufacturers (OEMs) for vehicle specifications. The vehicle-data provided by OEMs is unstructured, usually available in the form of cumbersome multiple PDF documents. Earlier, it took about 2 weeks to manually convert the data into a suitable form that can go live on the website.
- Having an updated Edmunds’ database and a real-time inventory that accurately describes Vehicle Identification Numbers (VINs).
- Speeding and augmenting the existing manual process of converting OEM data to live information that is available on their website (to reduce overall time-to-market).
Edmunds approached Silicon Valley Data Science (SVDS), a consulting firm specializing in agile and business-focus solutions, for help. SVDS built an extraction service that extracts and converts complex PDF data into raw text. Further, they built an “ontology-based attribute prediction engine using Idibon’s cloud-based Natural Language Processing (NLP) services”. This set-up allowed Edmunds to automatically extract data (information) about hundreds of car-relevant features and functions from the raw text. Post this, the NLP system could now create groups, categories, features-list and vehicle attributes to define and describe new car models in Edmunds’ database. The system now is thoroughly capable of handling all user queries and offers them with relevant search results and up-to-date information and specifications on hundreds of old and new cars.
- Edmunds can now use the NLP model not only for providing car information but for processing unstructured data from a variety of sources. It means more “analytics-muscle” and more business.
- The time-to-market reduced to 2 days from 2 weeks using this NLP data processing model.
- Reduced 95% of the backlogs at Edmunds’ end.
To read more about the contribution of data science in this whole process, please follow the link: http://bit.ly/2niFTkQ