Browse Prior Art Database

Personalized Image Tagging System

IP.com Disclosure Number: IPCOM000250384D
Publication Date: 2017-Jul-07
Document File: 3 page(s) / 153K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a solution that intelligently identifies the image sharing intent and provides personalized image tagging that reflects that intent. Personalized image tags help the image viewer understand the image content and the details referenced by the image producer.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

1

Personalized Image Tagging System

People share images to provide some additional details or specific or personal information. When sharing the image, the person might use different image characteristics to describe the main objects of that image. However, some image descriptions might not be that obvious for the image viewer. For example, color can be used to describe the referenced object in the image. If the image viewer is a colorblind person or the viewer uses the grey scale device for the image retrieval or viewing, then locating or identifying the object that referenced by color is difficult.

The novel solution is a method and system that intelligently identify the image sharing intent and provide personalized image tagging that reflects that intent. The proposed method helps the image viewer understand the intended image details.

The system uses Natural Language Understanding to recognize the intent of the image sharing and extract the entity object relevant to that intent. It applies image recognition to detect the described objects. The system generates the personalized image tags that reflect the user intent and disambiguates multiple visual objects of the same type by matching with the object visual properties extracted from the conversations and phrases

System Components

1. System Intent Classifier: trained to detect the image sharing intent. The classifier is trained on the phrases such as: “let me show you my house.”, “Here is the picture of my horse”, etc. The intent classifier is applied at runtime to detect the statements with the image sharing intent.

2. System Entity Extractor: extracts the image object entity from the image sharing intent phrases. It also extracts from the phrases the visual descriptions of the objects defined as visual entities or properties.

3. Visual Recognition Module: system applies visual recognition to detect the referenced object in the image. The system disambiguates multiple visual objects of the same type by matching with object visual properties provided by the system entity extractor.

4. System Image Renderer: generates and applies personalized tags to the image.

System Flow 1. System detects the statements with the image sharing intent by analyzing the

phrases, comments and conversation content. For example: A. The person writes the message: “Let me show the picture of my house. It

is the red brick house next to the grey building....