Deal Aggregator With NLP Engine

NLPProduct matchingWeb scrapingAmazon API

Automated deal platform powered by scraping and NLP

VidaPlayer is a leading videogames reseller in Spain, currently expanding into EU and LATAM regions. Along with its expansion, VidaPlayer is exploring affiliate marketing as a new revenue stream that would be synergic with its core e-commerce business. Bohemian AI partnered with VidaPlayer to develop an automated deal platform, www.vidaofertas.es.

Challenge

Online deal platforms are heavily dependent on human labour (community-driven websites) and/or vendor-specific API/data feed integrations (price comparison websites), both of which makes them difficult to scale. Both designs also inevitably lack a systematic approach to price monitoring.

Solution

Build an automated deal platform that automatically identifies top-selling products in given categories, discovers them on market’s leading e-shops, monitors their prices, identifies the best deals and publishes them on various channels, such as website and social media.

System Architecture

System architecture

Technologies

  • Most of the system is built using Python.
  • Scraping components use Scrapy framework.
  • Product matching engine uses a combination of pre-processing, term weighting, document vector similarity comparison and domain-specific rules. Experiments with neural networks have shown promising results and are likely to be implemented at larger scale in the next phase.
  • Internal UI components were built using Elm language.
  • All components are hosted on AWS infrastructure, and where possible, they run as serverless Lambda functions.

Results

10,000+

product pages

checked for price every hour

100-200

best deals

presented to the users

5-10

minutes

of human maintenance / day

Read More Case Studies