Document Ontology Extractor(DOE)

Research Team:Govind R Maddi, Jun Zhao Chakravarthi S Velvadapu

Faculty:Dr.Sadanand Srivastava

Dr.James Gil De Lamadrid

Joint Project of University of Maryland, Baltimore County

Bowie State University

Sponsored byDepartment Of Defense

OVERVIEW

1. The system takes text documents as its input

2. Performs semantic analysis on these documents

3. Generates useful ontology

4. Represents it graphically

GOAL

To build an Ontology utilizing

• Statistical methods

• A small amount of user feedback

• Automation

Normalization

Latent Semantic Indexing(SVD)

Pre-processing

Text Document

Document Ontology

Graph Construction

GUI

Architecture of DOE

INPUT

•Text documents

Pre-processing

Read-in text file

Extract meaningful terms

Count their frequencies

Normalization

Calculate weight of each term using

W i,k = frequency i,k nk

Σ frequency j,k

j=1

Calculate weight of each term using

W i,k = frequency i,k nk

Σ frequency j,k

Normalization(contd)

Calculate normalized weight using

W i,k w(i,k)

nk

sqrt(Σ w2(j,k))

j=1

Latent Semantic Indexing(LSI)

Statistical method representing documents by statistically independent concepts

Based on Singular Value Decomposition (SVD)

Singular Value Decomposition (SVD)

A technique that decomposes a given matrix into three components – U, S and V.

SVD (contd)

m x n term-document matrix A, of rank r, can be expressed as the product:

A = U * S * VT

U is m x r term matrix

S is r x r diagonal matrix

V is r x n document matrix

SVD (contd)

Diagonal of S contains singular values of A in the descending order.

SVD (contd)

A is formed from LSI as follows: A = US * SS * VsT

US - derived from U removing all but the s columns

SS - derived from S removing all but the largest s singular values

VsT - derived from VT removing all but the s corresponding rows

SVD (contd)

US

VsT

Am x n

Um x r

Sr x r

VT

r x n

SS

Document Ontology

Build Concept Nodes and Term Nodes

using the document matrix (V) and term matrix (U).

Building concept nodes from term matrix(U)

A concept node contains information about

• Concept name

• Terms that belong to that concept

• Respective weights of terms in that concept

Building concept nodes from term matrix(U) (contd)

Naming convention:• Generates automatically

• A hyphenated string of the five most high frequent terms in that concept

Building concept nodes from term matrix(U) (contd)

A concept node represents a document

Each column in U corresponds to a concept node

Building term nodes from term matrix(U)

A term node contains information about

• Term name

• Concepts to which it belongs

• Its respective weight in each concept

Building term nodes from term matrix(U) (contd)

Naming convention:• Generates automatically

• Simply named using the term name

Building term nodes from term matrix(U) (contd)

A term node represents a term

Each row in U corresponds to a term node

Graph Construction

A bipartite graph is constructed with concept nodes and term nodes

A concept node is connected to all term nodes that belong to it.

A term node is connected to all concept nodes to which it belongs.

Graph Construction (contd)

Term 1

Concept 1

Concept 2

Term 2

Term 3

Term 4

Term 5

Graphical User Interface

(GUI)

GUI (contd)

GUI consists of• Concepts list• Terms list• Display for bipartite graph• Display for list of files in ontology

GUI

To view terms related to a concept, user selects that concept from concepts list

To view concepts related to a term, user selects that term from terms list

GUI (contd)

To view only terms related to a specific concept:

• Select that concept from concepts list• Select checkbox “Display Selected Ones

Only”

Result:• GUI displays ONLY relations between

selected terms and concepts

GUI (contd)

To view only concepts related to a term:

• Select that term from terms list• Select checkbox “Display Selected Ones

Only”

Result:• GUI displays ONLY relations between

selected terms and concepts

GUI (contd)

To highlight relationship between a term and a concept:

• Select that term or concept from terms or concepts list

• Click on line connecting term and concept

GUI – File Operations

New

Open

Save

saveAs

Close

Exit

GUI – Ontology Updates

Add

Delete

ChangeSVDThreshold

changeConcThreshold

foldInDoc

defaultBuild

GUI – Ontology Updates

Add:• Click on Add

• Select file to be added from file chooser popup menu

• Choose whether to build now or not

• If yes document is added and displayed

• If no GUI remains unchanged

GUI – Ontology Updates

Delete:• Click on Delete• Select file to be deleted from file chooser

popup menu• Choose whether to build now or not• If yes document is deleted and displayed• If no GUI remains unchanged

GUI – Ontology Updates

changeSVDThreshold:• SVDThreshold controls the largest s

singular values that will be selected from S.

• Default value is 70% i.e. only the singular values higher than 70% of the highest singular value are selected

• User can change this default value

GUI – Ontology Updates

changeConcThreshold:• Controls the number of terms related to

a concept based upon term weight

• Default value is 70% i.e. only the terms with weights higher than 70% of the highest term weight are selected

• User can change this default value

GUI – Ontology Modifications

Rename• Renames a selected concept

DelTerm• Deletes a selected term

Undo• Ignores last modification and returns to the

previous state

Future Work

To investigate less expensive methods for adding new documents:

• Fold-In

• SVD update

Future Work

Fold-In:• A method to add new document(s) to

an existing ontology

• Uses the existing data in document addition process

• Less expensive process than the regular build method

Acknowledgements

We express our appreciation to

• Department Of Defense• University of Maryland, Baltimore County• Advisors, Bowie State University