KDD- Service based Numerical Entity Searcher (KSNES)

Presentation 2 on March 31st , 2009

Naga Sowjanya Karumuri

CIS 895 – MSE PROJECT

Naga Sowjanya Karumurisowji@ksu.edu

1

OUTLINE

¢ Project Data Flow Diagram¢ Action Items¢ Architectural Design¢ Test PlanFormal Inspection Checklist¢ Formal Inspection Checklist

¢ Project Plan¢ Prototype Demonstration¢ Questions / Comments

2

PROJECTDATA FLOWDIAGRAM:

NUMERICAL

ENTITY

SEARCHERSEARCHER

3

MODULES IN THE PROJECT

¢ Webpage (JSP): For requesting and receiving information from the service.

¢ POS Tagger (Java): Stanford POS Tagger

¢ Numerical Phrase Extractor (Java): Implemented ¢ Numerical Phrase Extractor (Java): Implemented using Shallow Parsing Technique

¢ Number-Unit/Date Pattern Recognizer (C++): Implemented based on the Numerical Quantifier developed by Benjamin Sapp, UIUC.

4

ACTION ITEMS

¢ Implemented Numerical Phrase Extractor

¢ Detailed Description of Test Plan

¢ Wrote Formal Specification using USE

¢ UML Representation of the System

5

ARCHITECTURAL DESIGN

Service Oriented Architecture

6

PACKAGE VIEW

Overall Package View

7

Class Descriptions, Attributes and Operations are contained in Architecture Design Document

SEQUENCE DIAGRAM

8

CLASS DIAGRAM(NPE PACKAGE)

9

CLASS DIAGRAM(NDPR PACKAGE)

10

IMPLEMENTING NUMERICAL PHRASE

EXTRACTOR

¢ Input: Tagged Text� I/PRP lost/VBD thirty-three/JJ dollars/NNS in/IN 1998/CD

¢ Regular Expressions are used to determine the numerical patterns in the input.numerical patterns in the input.� thirty-three/JJ dollars/NNS� in/IN 1998/CD

¢ Output: Numerical Phrases� thirty-three dollars� in 1998 11

TAGSET

12

SOME PATTERNS

¢ "\\d+-\\d+(/JJ|/CD) [a-zA-Z]+/NN"parses

\\d+-\\d+(/JJ|/CD) [a-zA-Z]+/NN

3-2/JJ lead/NN

20-20/JJ match/NN

¢ "(between|Between|from|From|In|in|since|Since|during|During)/IN ..../CD (([a-zA-Z]+/CC|[a-z]+/TO) ..../CD)?”

parses'between 1987 and 1997', 'in 2007 and 2008’ 13

20-20/JJ match/NN

ASSIGNING BOUNDS

¢ Words that will be detected so as to set the bounds like >, <, ~, =

¢ “ = ” is used if no words are mentioned

Bound Corresponding words

14

Bound Corresponding words

> more than, no less than, no fewer than, at most, over

< up to, not over, no more than, at least, less than, not over than

~ about, around, approximately, some, nearly, almost,

SOME PATTERNS

¢ [a-zA-Z0-9]+/CD( percent/NN)?( out/IN)? of/IN( the/DT)? ( [a-zA-Z]+/CD)?( [a-zA-Z]+/JJ)? [a-zA-Z]+(/NN|/NNS|/NNP)

parsesone of the five peopletwo of the groupsone of the rare cases89 percent of peoplefive of the seven former employees3 out of 5 people

15

PHRASES THAT CAN BE PARSED

Numerical Phrases

27 year-old boy

A 3-2 lead

9 in 10 people

About 100 miles per hour

200 adults and children

Temporal Phrases

Last year

Next week

Monday – Sunday

January–December

1956-60

16

$3 million

About two-thirds of the vote

The 17-mile drive

Less than 10% support

Six-bedroom apartment

5.987 ml

10:00 a.m. CST

From 400 to 500 miles

Mid-1990s

Between 1999 and 2008

17th centaury

18 April 2008

Dec 21, 2009

October 10th 1984

John, 67

Since 1998

PHRASES THAT ARE NOT CURRENTLY

PARSED

Numerical Phrases Temporal Phrases

six-pack of drinks 31st of March 1998

$100 more Since mid-November

252° (as POS can’t parse this) the January-April period

17

Future Work:

These phrases can also be parsed by adding more patterns to the current system but for now the most important and commonly occurring patterns are considered.

Current goal is to develop a basic idea of numerical phrase extraction.

FORMAL SPECIFICATION

¢ Created and validated using USE 2.3.1.¢ All Classes are specified

� All important attributes and methods are specified� Constructor methods are not specified

¢ Contained at the end of the Architectural Design ¢ Contained at the end of the Architectural Design Document

18

TEST PLAN

¢ Outputs are checked at each module by the developer by matching them to the results manually calculated� Check if the POS tagger has given the tagged text.� Check if the numerical phrases are extracted Check if the numerical phrase is explained to Value, � Check if the numerical phrase is explained to Value, Unit and Unit-Type.

¢ UML diagrams and the required specifications will be checked for consistency by two fellow MSE students

¢ User interaction will be tested by the developer and the technical inspectors. 19

FORMAL INSPECTION CHECKLIST

¢ The following items are to be checked:� The symbols used in the class diagram conform to UML

standards� The symbols used in the sequence diagrams conform to UML

standards� The classes in the class diagrams have corresponding

descriptions provided in the Architecture Documentdescriptions provided in the Architecture Document� The descriptions of the classes in the Architecture Document

are clear and concise� The classes in the USE model are consistent with those in the

Architecture Document� All the requirements in the Software Requirements

Specification have been covered in the Architecture Document� The multiplicities in the USE model have been depicted in the

class diagram20

PROJECT SCHEDULE

¢ Key Dates� Presentation 1: February 24th, 2009

¢ Complete Numerical Sub-Chunker

� Presentation 2: March 31st , 2009¢ Complete Numerical Phrase Extractor

� Presentation 3: April 10th, 2009� Presentation 3: April 10 , 2009¢ Patch up the modules¢ Develop a GUI¢ Set them up on the server

� To completely submit the documents by April 13th, 2009to the committee

� Final Portfolio submitted by April 15th , 2009

21

PROJECT SCHEDULE

22

PROTOTYPE DEMONSTRATION

¢ POS Tagger working� For now it works on the local machine

¢ Numerical Pattern Extractor¢ Numerical Pattern Extractor� For now it works on the local machine

23

PHASE 3 DELIVERABLES

¢ Action items¢ Component Design¢ Assessment Evaluation¢ Project EvaluationUser’s Manual¢ User’s Manual

¢ Formal Technical Inspection Checklists¢ Presentation 3¢ Executable Project ¢ Source Code

24

TO-DO LIST

¢ Revise the Documents¢ Revise Project Schedule¢ Work on the Phase3 deliverables¢ Final Demo

25

Questions??

Suggestions!!

THANK YOU 26