Product recommendation using sentiment and semantic analysis

Patent metadata

Field	Value
Publication number	US-9418375-B1
Application number	US-201615010646-A
Country	US
Kind code	B1
Filing date	Jan 29, 2016
Priority date	Sep 30, 2015
Publication date	Aug 16, 2016
Grant date	Aug 16, 2016

Abstract

In an approach to determine a product rating a computer receives a user request for a product rating. The computer retrieves from on-line sources, product information on the product and analyzes the product information to determine a first product rating. The analysis includes at least a sentiment, and a trend of the sentiment. The approach includes a computer identifying products similar to the product and retrieving from on-line sources product information on similar products. A computer extracts comments on the product from the similar product information and determines an adjustment to the first product rating based on an analysis of the comments and references to the product in the similar product information. The adjustment to the first product rating includes using a sentiment, a trend of the sentiment over time, and a frequency of comments and references to the product over time in the retrieved plurality of similar product information.

First claim

What is claimed is:

1. A method for providing one or more product ratings, the method comprising:
receiving, by one or more computers, a user request for a product rating for a product;
retrieving, by one or more computers, from a plurality of on-line sources, a first plurality of product information on the product;
the first plurality of product information including ratings, reviews, articles, and comments in blogs, websites, and social media sites;
determining, by one or more computers, a first product rating, based, at least in part, on an analysis of the first plurality of product information, the analysis of the first plurality of product information including at least, a sentiment, a trend of a sentiment over time, and a frequency of a plurality of comments and references to the product over time, and wherein determining the first product rating includes normalizing, by one or more computers, one or more product ratings extracted from the plurality of product information, wherein normalizing the one or more product ratings includes equating the ratings from one or more rating systems;
identifying, by one or more computers, one or more similar products to the product, wherein identifying the one or more similar products to the product comprises:
identifying, by one or more computers, one or more of product attributes, product characteristics, and a product type of the first plurality of product information, based on utilizing one or more of the following techniques: embedded metadata, tagging, natural language processing (NLP), deep linguistic processing, and software algorithms utilizing machine learning; and
sending, by one or more computers, at least one of the one or more product attributes, the product characteristics, and the product type to a content retrieving module to identify one or more similar products with the same product attributes, product characteristics, and product type;
retrieving, by one or more computers, from the plurality of on-line sources a second plurality of product information for the one or more similar products, the second plurality of product information on the one or more similar products include product reviews, ratings, articles, product specifications, and comments for the one or more similar products;
extracting, by one or more computers, a plurality of comments and references to the product in the retrieved second plurality of product information for the one or more similar products;
determining, by one or more computers, an adjustment to the first product rating based on the plurality of comments and references to the product in the retrieved second plurality of similar product information, wherein the determining includes determining at least one of: a sentiment, a trend of a sentiment over time, and a frequency of the plurality of comments and references to the product over time, and adjusting the first product rating based, at least in part, on one or more of: a number of the extracted plurality of comments and references to the product in the second plurality of product information on the one or more similar products and a change of sentiment over a period of time;
applying, by one or more computers, the adjustment to the first product rating to create a second product rating; and
sending, by one or more computers, the first product rating and the second product rating to the user;
the first product rating and the second product rating include a time stamp that is used to track the first product rating and the second product rating as changing over time;
displaying the first product rating and the second product rating as a function of time.

Assignees

• IBM

Inventors

• Cunico Hernan A
• Silva Asima

Classifications

• G06F16/3344

  using natural language analysis · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/951

  Indexing; Web crawling techniques · CPC title

• G06Q30/0282 Primary

  Rating or review of business operators or products · CPC title