Browse Prior Art Database

Interpreting Complex Graphs for optimal ingestion

IP.com Disclosure Number: IPCOM000241711D
Publication Date: 2015-May-26
Document File: 2 page(s) / 116K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method for interpreting complex graphs for optimal ingestion into a system capable of answering questions is disclosed.

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

Page 01 of 2

Interpreting Complex Graphs for optimal ingestion

Disclosed is a method for interpreting complex graphs for optimal ingestion into a system capable of answering questions.

Often, when ingesting documents into a system capable of answering questions, there exist situations where complex graphs are difficult for the system to interpret and understand. Usually, the graphs have detailed information based on the content which

would help the reader and system to identify an optimal answer.

There is a need for understanding complex/unstructured graphs in a document. The disclosed method interprets graphs in an optimal way so to be ingested by the system capable of answering questions and interpret the context of the graph to determine other facts that are not explicitly described in the content. This is done by determining the context and detailed information of each object within the graph and interpreting the relationships between objects.

Many different forms to depict information may be used. For example a first chart may depicts the billionaires in each country using a bar chart form. A second chart may 2 depicts the number of billionaires in the world using a complex/unstructured chart with color coded circles, circle sizes indicating the number of billionaires, and etc.. A third

chart may depicts a cartogram of number of billionaires in each state, which is slightly

structured yet complex.

The following is an example of an embodiment of the disclosed method:

System identifies all objects and their type (text, image, media, etc.) in the graph


1.

and registers them accordingly. System can leverage existing image processing prior art such as object segmentation to identify specific objects, and distance transforms to identify and separate overlapping objects (for example the text and the objects that are contained in charts 2 and chart 3.

System then determines the relationship between each object based on the context


2.

of each object. For example:

Text objects in chart 2 may be related by country names and the other text sets may be related by a numerical value.

Circle objec...