Intent prediction based recommendation system using data combined from multiple channels

Patent metadata

Field	Value
Publication number	US-9652797-B2
Application number	US-201414154120-A
Country	US
Kind code	B2
Filing date	Jan 13, 2014
Priority date	Jan 18, 2013
Publication date	May 16, 2017
Grant date	May 16, 2017

Abstract

User intent is identified while the user browses online and recommendations are provided to the user. The recommendations are based on the identified intent, interests, and preferences of the user who is performing the searches. The determination of user intent and interests is based on a statistical model derived from data compiled from the user and a plurality of other users. Other resources may also be determined to be relevant, for example, because of past interactions of the user, memberships of the user in ecommerce websites, the user's interests and preferences are similar to those of other users, and so on. The result of the user search is a ranked set of recommendations that is provided to the user.

First claim

The invention claimed is:

1. A computer implemented method for making recommendations to a user, comprising:
providing a processor configured for receiving a query from a user through a plurality of channels, said plurality of channels comprising at least two of the group of Web, mobile, interactive voice response (IVR), phone, automotive, and television;
during a data fusion stage, said processor configured for using said customer's Web journey, multi-channel journeys, and activities that generate data from multi-channels to match the query received from the user with a previous query from any of said user and one or more other users;
said processor configured for identifying a set of resources relevant to the received query and the previous query;
said processor configured for combining data from said plurality of channels in real time to predict user intent in connection with said identified resources, wherein unique identifiers are created, captured, and/or passed between said plurality of channels to identify and tag the user and the user's context uniquely for any of history, past behavior, steps progressed, obstacles, and issues encountered;
during an intent prediction stage, said processor configured for predicting in real time an outcome of a future selection by said user from a sequence of previous selections made by any of said user and from one or more frequent sequences of previous selections made by a plurality of users, said intent prediction being determined by application of an intent prediction function p 1 , comprising:
p 1 =Prob(VisitorεIntent 1)= f (Navigation, Customer Data, Geo);

said processor configured for using activities of a group of similar users to provide a basis for determining a current user's intent where insufficient information is available about said user or said user is a new visitor;
during a recommendation stage, said processor configured for collecting and analyzing information gathered during previous visits of one or more users, including that of a current user, to determine in descending order a most frequent sequence of selections;
said processor configured for ranking results of said predicting based on said user intent by application of a function comprising:
Prob(Item 2|Item−1)=Σ p i *f i (features),

where interest in a product comprises a weighted sum of interest in each intent class I expressed as f i (features) and weightage comprises the probability of a user's intent being that of a specific intent class i expressed as p;
said processor configured for presenting a ranked order of recommended results to said user;
said processor configured for displaying products or services having the highest scores through any of a plurality of channels, wherein each channel has a different setting for top products or services;
said processor configured for proactively identifying a user need and initiating user contact and/or interaction through any of chat sessions and telephonic contact with a call center;
said processor configured for, during user interaction, determining a specific type of user engagement mode based upon the intent of the user; and
said processor configured for providing a recommendation to said user based on any of information gathered through said contact and/or interaction and past user activities.

2. The method of claim 1 , further comprising:
applying user preferences in connection with determining and ranking said recommendations.

3. The method of claim 1 , further comprising:
applying user related information in connection with determining and ranking said recommendations.

4. The method of claim 3 , wherein said user related information comprises any of past interactions, previous purchase history, geographical location, time of day, client device used, operating system, and a current browsing session.

5. The method of claim 1 , further comprising:
said user interacting with a product or service provider in connection with any of said query and said recommendations.

6. The method of claim 1 , further comprising:
said processor configured for implementing an intent module for receiving a query from said user;
said intent module matching said received query to a previous query to identify a set of resources that are relevant to both said received query and to each of the items in said set of resources;
said intent module compiling information concerning prior user search behavior; and
said intent module inferring said user's interests and intent from said information.

7. The method of claim 6 , further comprising:
said intent module presenting more relevant search results to subsequent users who are executing a same or a similar search query.

8. The method of claim 1 , further comprising;
collecting said information from a plurality of users and compiling or aggregating said information to provide a statistical model.

9. The method of claim 1 , further comprising:
said processor configured for implementing a monitoring module for tracking user activities.

10. The method of claim 1 , further comprising:
said intent prediction module recording each selection in said sequence of prior selections and adding a most recent selection to said sequence of prior selections that define a journey taken by said user when said user initiates a series of selections online;
said intent prediction module comparing a journey or sequence currently selected to the most frequent sequences of selections made by a plurality of users; and
said intent prediction module predicting a current intent of said user therefrom;
wherein relevant information is displayed or other actions are undertaken to assist said user in making additional or future selections that are necessary to achieve a desired end result.

11. The method of claim 1 , further comprising:
identifying user preferences based upon user-related information, wherein said user-related information comprises any of past interactions, purchase history, geographical location, time of day, client device used, ands operating system.

12. The method of claim 1 , further comprising:
considering said user's original intent and preferences on a decision level; and
considering users who exhibit similar intent on an aggregate level to have similar preferences to said user.

13. The method of claim 1 , further comprising:
said intent prediction module learning user preferences; and
said intent prediction module assigning a score or rating once said preferences are learned.

14. The method of claim 1 , further comprising:
based on said user's exhibited behavior, adjusting a weightage assigned to one or more specific product or service features to be high when compared to other product or service features.

15. The method of claim 1 , further comprising:
generating a score for user preferences by any of a ranking procedure and statistical functions;
assigning said user different weights for each of one or more product or service features, wherein all of the weights considered together comprise a preference.

16. The method of claim 15 , further comprising:
aggregating said weights for user preferences by either of multiplication or addition of each weight into a single score for a specific product or service.

17. The method of claim 14 , further comprising:
adjusting a default preference to be in line with said user's exhibited behavior as more information about user interaction is known.

18. An apparatus for making recommendations to a user, comprising:
a processor receiving a query from a user

Assignees

• 24/7 Customer Inc

Inventors

• Vijayaraghavan Ravi
• Kulkarni Subhash Ramchandra
• Adusumilli Kranthi Mitra

Classifications

• G06Q30/0631 Primary

  Recommending goods or services · CPC title

• G06Q30/0201

  Market modelling; Market analysis; Collecting market data · CPC title

• G06F16/951

  Indexing; Web crawling techniques · CPC title

• G06Q30/02

  Marketing; Price estimation or determination; Fundraising · CPC title