knowledge modelling assignment 2 (msc)jessicac/project/kmm/commokads example 2009... · this...

35
Knowledge Modelling and Management Assignment 2 Sample Solution 27 March 2009

Upload: hatuong

Post on 06-May-2018

216 views

Category:

Documents


1 download

TRANSCRIPT

Knowledge Modelling and Management Assignment 2

Sample Solution

27 March 2009

Contents

1 Organizational Context Analysis 1

1.1 Organisation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.2 Problems and Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.3 Potential Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.4 Variant Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.5 Process Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1.6 Knowledge Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2 Feasibility (Worksheet OM-5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2.1 Business Feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2.2 Technical Feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2.3 Project Feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2.4 Proposed Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Domain Model 11

2.1 Domain Ontology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1.1 Justification of the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Task Models 15

3.1 Task Models TM-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1.1 Receive Request . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1.2 Elaborate Request . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1.3 Search Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.1.4 Rank Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1.5 E-mail/Print Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1.6 Explain Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1.7 Bill Client . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1.8 Task Time Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.2 Static Information Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.3 Dynamic View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.4 Task Model TM-2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

i

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

4 Organizational Task Analysis 244.1 Impacts and Changes in Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.2 Task-specific impacts and changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264.3 Attitudes and Commitments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264.4 Proposed Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5 System Design Document 285.1 Use Case Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285.2 Proposed software architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

5.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.2.2 Subsystem decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.2.3 Hardware/software mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.2.4 Persistent data management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305.2.5 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

6 References 32

ii

Chapter 1

Organizational Context Analysis

1.1 Organisation Model

This section describes the model of the PPRC organisation, following the methodology of the Com-monKADS knowledge engineering and management process. During the creation of this context analysis,the appropriate CommonKADS worksheets were completed, and are reproduced and referenced here.

1.1.1 Background

The client is a small business operating in Edinburgh with five full-time employees. The client’s primarybusiness is to research properties for their customers, based on the customer’s unique requirements. Thecompany does not own any properties, but instead performs research on local property market web sites tofind relevant results for their customers. The company’s target market includes both corporate customersand individual customers. Currently, the company uses several web sites to make their recommendations,and the process is done by hand. Because of increased business opportunities, the client wishes to improvethe speed with which they can provide recommendations, and the quality of those recommendations.

1.1.2 Problems and Opportunities

The organisation’s problems and opportunities were analysed according to the CommonKADS methodol-ogy, and worksheet OM-1 was completed. The worksheet appears in Table 1.1. Several problems wereidentified with the current organisational structure and process. Chief among these is that the client turnsaway about 30% of potential business because of lack of capacity. The capacity constraints stem mostlyfrom the amount of time that it takes to handle a request, up to 80 minutes in some cases (discussed furtherin the task analysis, later in this document). Other problems include the extra difficulty in researchingoverseas properties, and that the client is not able to provide the quantity of results that they hope to, againdue to time constraints. These problems are minor, however, because the volume of business for overseasproperties is low, and increasing the number of results for an enquiry has not been shown to influence thecompany’s turnover.

1

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

1.1.3 Potential Solution

The proposed solution involves constructing a knowledge-based system to assist with the processing andhandling of requests, searching, and ranking of results. The system would automate the reception andassignment of requests, be capable of retrieving and classifying appropriate properties, and producing apreliminary ranking and report.

Table 1.1: Worksheet OM-1: Problems and opportunities

Organisation Model Problems and Opportunities Worksheet OM-1

PROBLEMS AND

OPPORTUNITIES• Demand has recently grown tremendously

• 30% of potential customers turned away because of slow re-sponses or not being able to respond

• Assessment for individual clients takes too long

• Searching and ranking overseas properties takes too long

• Can only offer a limited number of properties to a client request(top 3) because of time constraints

ORGANISATIONAL

CONTEXTMission: Search properties and make recommendations based on client

requirements and deliver high quality, timely information.

External Actors: Property web sites, customers

Strategy: Provide high-quality information on properties in a timelymanner.

SOLUTION I: Hire more staff to handle peak demand, or II: Develop an automatedsystem to produce high-quality results quickly with minimal interven-tion; introduce web-based tracking system for client requests to manageinternal workflow and allow external clients to track request progress;implement a training programme to improve the knowledge of less-skilled agents.

1.1.4 Variant Aspects

An analysis of the company’s structure and purpose was carried out, and is documented in Worksheet OM-2, which appears in Table 1.2. A process model of the current workflow was also crated, and is shownin Figure 1.1. A challenge to automating the business process is that the client uses its knowledge aboutthe property industry to differentiate itself from competitors, but this tacit knowledge may be difficult toextract and formalise.

2

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 1.2: Worksheet OM-2: Variant Aspects

OrganizationalContext

Worksheet OM-2 Variant Aspects

STRUCTURE The company has a flat organizational structure with few formal roles.Everyone does a bit of everything, although there is one owner-managerin charge.

PROCESS See Figure 1.1PEOPLE There are five Agents in the company, who work in the process de-

scribed above.RESOURCES Primarily consist of web sites for property search. May also include a

client database.KNOWLEDGE Includes: Assessment criteria for property suitability and relevance; Ur-

gency rules for processing incoming requests; Knowledge of the prop-erty industry rules and local areas.

CULTURE & POWER Flat organizational structure; employees are currently working at ornear capacity.

3

KM

M:P

RIN

CIPA

L.PR

OP

ER

TY.R

EC

OM

ME

ND

ER.C

OM

(MSc)

Sample

Solution

Search Properties Rank Results Report Results

Explain Recommendations

X

Bill Customer

Receive Request Elaborate Request X

E-mail Results

X

X

Print Results

Figure 1.1: Current process model.

4

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

1.1.5 Process Breakdown

An analysis of the business process was carried out, and is documented in Worksheet OM-3, Table 1.3.The table should be read with reference to Figure 1.1. A brief description of each process appears below:

Receive Request Receive a request from a client, and assign it to an available agent. When multiplerequests arrive at once, prioritise them according to urgency rules.

Elaborate Request Clarify the request with the client; obtain additional information about requirementswhere necessary.

Search Properties Use the web sites to search for relevant properties.

Rank Results Rank the results from the web search.

E-mail/Print Results Print the results, or e-mail them to the client.

Explain Recommendations Conference with the client to discuss the recommendations.

Bill Client Invoice the client. This task was not explicitly discussed in the client report, but is inferredfrom the nature of a business.

The tasks 3, 4, and 6 are the most knowledge-intensive of all of the tasks. As will be seen whenWorksheet OM-4 is discussed, these tasks share common knowledge requirements.

5

KM

M:P

RIN

CIPA

L.PR

OP

ER

TY.R

EC

OM

ME

ND

ER.C

OM

(MSc)

Sample

Solution

Table 1.3: Worksheet OM-3: Process BreakdownTask Performed By Where Knowledge Asset Intensive? Significance

1 Receive Request Agent Property Recommendations Urgency rules for re-quest assignment

Yes 2

2 Elaborate Request Agent Property Recommendations Knowledge ofamenities and elicita-tion questions

Yes 2

3 Search Properties Agent Property Recommendations Knowledge of areasand amenities, prop-erty relevance rules,web site operation

Yes 5

4 Rank Results Agent Property Recommendations Knowledge of clientneeds, areas andamenities, relevancerules

Yes 5

5 E-mail/Print Results Agent Property Recommendations No 16 Explain Recommendations Agent Property Recommendations Knowledge of rec-

ommendations &property amenities

Yes 4

7 Bill Client Owner/Manager Management No 3

6

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

1.1.6 Knowledge Assets

An analysis of the knowledge assets of the organization was carried out, and is shown in Table 1.4. Theknowledge of the property industry is the knowledge asset that is used the most in the organization’s tasks.None of the knowledge assets is currently in the right form, and the quality needs to be improved of allof the knowledge assets. Improving this knowledge quality, and the accessibility of the knowledge couldimprove the organization’s efficiency.

7

KM

M:P

RIN

CIPA

L.PR

OP

ER

TY.R

EC

OM

ME

ND

ER.C

OM

(MSc)

Sample

Solution

Table 1.4: Worksheet OM-4: Knowledge Assets

Knowledge asset Possessed by Used in Right form? Right place? Right time? Right quality?

Urgency rules forordering clientrequests

Agent 1 No, undocumented→ electronic

Yes Yes No, not formal

Assessment criteriafor propertyrelevance

Agent 3, 4 No, undocumented→ electronic

Yes No No, not formal

Knowledge of theproperty industry,property amenities,areas of town, etc.

Agent 2, 3, 4, 6 No, undocumented→ electronic

Yes Yes No, not formal,varies from agent-to-agent, and relieson information fromexternal sources

8

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

1.2 Feasibility (Worksheet OM-5)

This section documents the feasibility of the proposed solution.

1.2.1 Business Feasibility

The organization will benefit from the proposed solution by increasing its capacity to process requests.Because the system will partially automate the most knowledge-intensive tasks, less agent intervention willbe required for each request, which should decrease the processing time. It is expected that the system willsave about 70% of the time spent on search and ranking, and free up agents to spend more time consultingwith customers and soliciting new business. It is estimated that the time savings due to this system wouldequate to an approximate 30% increase in business, yielding additional revenue of approximately £45,000per year (based on an estimated revenue of £150,000, which is based on a staff of 4 each being paid anaverage of £25,000, and a manager earning about £35,000).

The system is estimated to cost approximately £25, 000 to implement, including consulting time,development, testing, deployment, training, and equipment purchases. Furthermore, the system will have arecurring annual maintenance requirement of approximately £2, 000. Such a system, therefore, would havea simple payback period of under 8 months.

The economic and business risks to implementing such a system include the potential for asudden economic downturn or change in the property market which would reduce the amount of businessavailable to the client. In such a case, the benefit of installing the system would decrease, and the paybackperiod could increase significantly.

1.2.2 Technical Feasibility

The system is not required to perform complex inferences over the knowledge domain. Most propertiescan be easily classified, and similar tasks have been automated previously. There are numerous systems inexistence that can perform the types of reasoning and inference that this system will require. Because oftime and budget concerns, the system will need to be produced within six months of the contract date. Thesystem will need to be able to produce results of similar quality to an inexperienced agent, 95% of the time.Also, because the system will become critical to the operation of the company, it must be available 99.99%of the time during regular 8:00 AM to 5:00 PM business hours, excluding planned maintenance periods.To this end, the system will be quality tested according to generally-accepted software engineering andsoftware testing practices.

The users of the system will be specially-trained, and thus the requirements for an extremelysophisticated user interface are diminished. A web-based UI will be required to process incoming clientrequests via the Internet. Because the system will perform automated ranking and searching of externalproperty web sites, the system will need to interact with those sites. Neither of the web sites that the clientcurrently uses has a web services based API. Therefore, the system will need to implement an interface thatcommunicates via standard HTTP form submission, and reads the resulting data. This approach is fragile,and extremely prone to breaking. The fragility of this interface is accounted for in the maintenance cost ofthe tool. It may be possible to connect to a centralised property listing service that provides an EDI or webservice interface.

9

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

1.2.3 Project Feasibility

The client has expressed interest in implementing a system to assist with their recommendation process;the project has buy-in from both the management and employees of the organization. The client has agreedto commit the necessary time, budget, and equipment to the project. The company that has been selectedto implement the project has extensive experience implementing knowledge-based systems, and is able toleverage existing designs and code for this project. The project is therefore feasible along these measures.

1.2.4 Proposed Actions

The following recommended actions should be taken:

1. The project focus should be on capturing the knowledge required to automate the business processespreviously identified as being the most knowledge-intensive, and then constructing a system to par-tially automate these process tasks.

2. For this focus area, the solution direction would be to interview employees to attempt to capturedomain knowledge and process knowledge that they use to make rankings and decisions, and encap-sulate them in an ontology.

3. The result will be a system that is capable of supporting the agents in making property recommen-dations, at the cost and benefits previously discussed.

4. The next actions would be to further investigate the tasks identified in OM-3, and determine the taskswhich are most suitable to automation.

5. Reconsideration would be appropriate in case there is a change in business conditions that wouldobviate the need for this system.

10

Chapter 2

Domain Model

2.1 Domain Ontology

A graphical domain model was produced, and is shown in Figure 2.1. The model was constructed basedon an understanding of the domain from the description in the report provided by the client. Addi-tional information was obtained from the web sites http://www.espc.co.uk/ and http://www.rightmove.co.uk/.

The domain model contains approximately 23 classes and three individuals, and documents therelations between them. Descriptions of the classes and properties are given in Table 2.1.

2.1.1 Justification of the Model

The model was constructed based on an analysis of the domain as described in the client report, and reviewof the two aforementioned web sites. The model was constructed to capture the most important criterafor deciding on the suitability of a property for a client. The hierarchical nature of the ontology allowsproperties to be classified and retrieved based on a number of criteria, including:

• Whether the property is in the UK or overseas;

• The types of amenities available in the neighbourhood of the property;

• Whether the property is available for purchase, rental, or holiday rental;

• The features of the property.

These are the criteria that the client’s report describe as the most important measures of a prop-erty’s suitability. This model will allow a knowledge-based system to classify properties based on theirdescriptions from an external web site, according to the most relevant criteria from a client request.

11

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Thing

:forSaleType

:toLetType

:holidayLetType

Location

UKLocation

OverseasLocation

Property

House

Detached

SemiDetached

Flat

PropertyFeature Balcony

Garden

Parking

StreetParking

GarageParking

NeighbourhoodAmenity

EducationalFacility

UniversitySchool

PrimarySchool

SecondarySchool

PropertyType

Pubs

Park

DogPark

ChildrenSwingPark

PublicTransportation

Shops

has-amenity

in-location

contains-property

has-feature

1

*

1

*

Legend

is-a relation

has-a relation

Class

hasPropertyType

Figure 2.1: A graphical representation of the domain model.

12

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 2.1: Descriptions of the classes, individuals, and properties of thedomain ontology

Entity Name Description

Classes

Thing The parent class of all classes in the ontology.NeighbourhoodAmenity Some feature of a neighbourhood, like a shop, or proximity to public

transportation.Location Some geographical location that contains properties, and has amenities

to distinguish it from other geographical locations.Property An enclosed space designed for residential living.PropertyType The contract type under which the property is available.PropertyFeature A component of a property that can distinguish it from other properties,

and make it more desirable.UKLocation A Location inside the United Kingdom.OverseasLocation A Location outwith the United Kingdom.House A Property with a street entrance and multiple storeys.Flat A Property inside a building which contains other Flats.Pubs A social place in which drinks and sometimes food are served.Park A public space, usually with grass and trees.PublicTransportation A public system for moving people around the city.Shops A place to do shopping, may include grocery stores.Detached A type of house that is not connected to any other house.SemiDetached A type of house that shares one or more walls with another house.Balcony A fenced outdoor space that is not at ground floor or on the roof.Garden An outdoor space at ground that is used for growing plants.Parking A place to store a vehicle when not in use.StreetParking Overnight parking on the street, reserved for residents.GarageParking Secure overnight parking in an enclosed space.DogPark A Park that is set aside for dog owners, with designated off-leash zones.ChildrenSwingPark A Park with facilities for chilren to play on swings and climbing gyms.EducationalFacility A place where students learn.PrimarySchool An EducationalFacility for children aged 6–13.SecondarySchool An EducationalFacility for children aged 14–18.UniversitySchool An EducationalFacility for higher learning.

Individuals

:forSaleType The property is for sale.:toLetType The property is for rent.:holidayLetType The property is for holiday rental only. This type applies only to prop-

erties with an OverseasLocation.

Relations

13

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 2.1: Descriptions of the classes, individuals, and properties of thedomain ontology (continued)

Entity Name Description

in-location A Property exists inside a location. A Property may only have oneLocation. Inverse of contains-property

contains-property A Location encloses one or more properties. This is the inverse of in-

location.has-feature A property has zero or more Features.hasPropertyType A Property has as its type one of the instances of PropertyType

Properties (not shown in diagram)

Property:monthlyCost(nonNegativeInteger)

The monthly rental cost of the property.

Property:purchaseCost(nonNegativeInteger)

The purchase price of the property.

Property:numberOfBedrooms(nonNegativeInteger)

The number of bedrooms in the property

14

Chapter 3

Task Models

3.1 Task Models TM-1

3.1.1 Receive Request

Task This is task 1, Receive Request.

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value The task receives a request from a client, evaluates it, and assigns it to an agent for furtherwork. In case multiple requests arrive at the same time, this task prioritizes the work.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled Input Objects The system receives a request from a customer, containing informationabout the relevant features that the customer is looking for. Output Objects An assigned request forfurther processing.

Timing and Control This task occurs for every instance of the process. It is estimated to take no morethan 2–3 minutes per request.

Agents A property agent is responsible for the task.

Knowledge and Competence The urgency determination is knowledge intensive and requires knowledgeof the rules for determining urgency. Assignment of requests to the right individual may also requireknowledge of staff expertise.

Resources This task should consume no more than 5 minutes.

Quality and Performance The task executes successfully if a request has been assigned to an agent.

3.1.2 Elaborate Request

Task This is task 2, Receive Request.

15

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value This task examines a request from a client and determines what gaps, if any, need to beaddressed before the request can be successfully serviced.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled The input object is a request that has been assigned from the previous task. The outputobject is an elaborated request.

Timing and Control See the timing analysis in Table 3.1.

Agents A property agent is responsible for the task.

Knowledge and Competence Knowledge of amenities in the local areas, style of neighbourhood, andappropriate elicitation questions.

Resources This task consumes about 9 minutes for a corporate client, but about 50% more for an individualclient.

Quality and Performance A successful execution of this task occurs when an elaborated request is pro-duced that contains enough information to successfully generate search parameters.

3.1.3 Search Properties

Task This is Task 3, Search Properties.

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value This task examines an elaborated request, and performs a search of property web sites forrelevant properties.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled This task takes an elaborated request, and outputs a listing of properties that match theclient requirements to some degree.

Timing and Control See the timing analysis in Table 3.1.

Agents A property agent is responsible for the task.

Knowledge and Competence Requires knowledge of areas and amenities, property relevance rules, andoperation of the property search web sites.

Resources Property search web sites, primarily, and approximately 10–40 minutes of staff time.

Quality and Performance The task has executed successfully if it produces relevant search results.

16

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

3.1.4 Rank Results

Task This is task 4, Rank Results.

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value The goal of this task is to produce rankings for the search results that are appropriate tothe client’s requirements.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled This task has as an input the search results from Task 3, and produces a ranked list ofproperties.

Timing and Control See the timing analysis in Table 3.1.

Agents A property agent is responsible for the task.

Knowledge and Competence This task requires knowledge of the client’s needs, the areas and amenitiesof a neighbourhood, and the rules for sorting properties by relevance.

Resources This task requires the search results, and may require access to the property search web sites.

Quality and Performance This task has executed successfully if it produces a ranked list of results.

3.1.5 E-mail/Print Results

Task This is Task 5, E-mail/Print Results.

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value The purpose of this task is to produce a copy of the ranked results suitable for distributionto the customer.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled This task takes a ranked list of results, and prints it or e-mails it to the client, producinga printout or e-mail message.

Timing and Control See the timing analysis in Table 3.1.

Agents A property agent is responsible for the task.

Knowledge and Competence This task does not require any knowledge.

Resources This task requires access to a computer, and possibly a printer.

Quality and Performance This task has executed successfully if a printout is produced or an e-mail issent to a client.

17

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

3.1.6 Explain Recommendations

Task This is task 6, Explain Recommendations.

Organization This is performed by the the Property Recommendation organization by persons in theAgent role.

Goal and Value The purpose of this task is to ensure that the customer understands the proposed rec-ommendations, and has the ability to comment and provide feedback to ensure that they meet hisneeds.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled This task takes a printout or e-mail, and produces a verbal explanation to the client onthe properties and why they have been selected.

Timing and Control See the timing analysis in Table 3.1.

Agents A property agent is responsible for the task.

Knowledge and Competence This task requires knowledge of the neighbourhoods and amenities, theranking rules, and the customer’s needs.

Resources This task requires the printout produced in Task 5, and may require access to the property websites.

Quality and Performance This task has successfully executed if the client is satisfied with the resultsproduced. In case the client is dissatisfied, the process returns to the Search task.

3.1.7 Bill Client

Note: This task is not strictly part of the recommendation workflow, and thus will not be discussed further

in this document. It is included here for the sake of completeness.

Task This is Task 6, Bill Client

Organization This is performed by the the Management organization by persons in the Owner/Managerrole.

Goal and Value This task bills the client for the work performed.

Dependency and Flow Refer to the as-is process model in Figure 1.1 on page 4.

Objects Handled This task takes a confirmation of client satisfaction as input, and produces an invoice asoutput.

Timing and Control This task is expected to take no more that 2–3 minutes.

Agents An owner/manager is responsible for the task.

Knowledge and Competence This task requires knowledge of the billing rules for the customer.

Resources This task requires access to a billing system.

Quality and Performance This task has executed successfully if a bill has been produced for the client.

18

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 3.1: Task time analysis by client type. Times shown are worst-case. Total shows range for best-caseto worst-case, based on search, rank, and report processing time of 15–60 minutes.

Corporate IndividualPercent Estimated (minutes) Percent Estimated (minutes)

Elaborate Request 13 10 18 15Search Properties 50 38 47 38Rank Results 20 15 19 15Report Results 8 6 8 6Print or E-mail Results 2 2 2 2Explain Recommendations 7 5 7 5

Total 30–75 35–80

3.1.8 Task Time Analysis

Table 3.1 shows an analysis of the time each task takes. The timings shown in the Estimated column arebased on a worst-case scenario of 60 minutes to service a search-rank-report sequence, yielding an overall75 minute process time for a corporate client, or 80 minutes for an individual. The increased time for anindividual is due to the increased elaboration time required.

3.2 Static Information Structure

A structure of the objects involved in the process appears in Figure 3.1, and described below:

Request The initial non-elaborated client request.

Elaborated Request A request that has had additional information filled in, based on a conference withthe client.

Search Results A set of results from different property web sites that may match the client’s needs.

Ranked Results An ordered list of results from the search, based on the client’s criteria.

Printed Results A hard copy of the results for presentation to the client.

3.3 Dynamic View

A dynamic view of the process model is shown in Figure 3.2. The model shows the objects that areproduced at each stage of the workflow.

3.4 Task Model TM-2

Worksheet TM-2 were completed for each knowledge asset identified in the Worksheet OM-4. The taskworksheets appear in Tables 3.2–3.4.

19

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Request

Elaborated Request

Search Results

Ranked Results

Printed Results

Figure 3.1: The static structure for the process, showing the objects involved in the workflow.

Receive Request

Elaborate Request

SearchRank ResultsReport Results

Print Report Explain Recommendations Bill Client

:Request :ElaboratedRequest

:SearchResults

:RankedResults

:Report

:PrintedResults

Figure 3.2: Dynamic view of the workflow, including objects produced at each task transition.

20

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 3.2: Task Model TM-2 for urgency rules for ordering client requests

NAME Urgency rules for ordering client requestsPOSSESSED BY Property agentUSED IN Task 1DOMAIN Planning and organization

Nature of the Knowledge Bottleneck/to be improved?

Formal, Rigorous XEmpirical, QuantitativeHeuristic, Rule of thumb XHighly specialized, domain-specificExperience-based X XAction-based XIncompleteUncertain, may be incorrectQuickly changingHard to verify XTacit, hard to transfer X X

Form of the knowledge

Mind X XPaperElectronicAction skill X XOther

Availability of knowledge

Limitations in timeLimitations in spaceLimitations in access X XLimitations in quality XLimitations in form

Remarks:

21

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 3.3: Task Model TM-2 for assessment criteria for property relevance

NAME Assessment criteria for property relevancePOSSESSED BY Property agentUSED IN 3,4DOMAIN Real estate

Nature of the Knowledge Bottleneck/to be improved?

Formal, Rigorous XEmpirical, QuantitativeHeuristic, Rule of thumb XHighly specialized, domain-specific XExperience-based X XAction-basedIncompleteUncertain, may be incorrectQuickly changingHard to verifyTacit, hard to transfer X X

Form of the knowledge

Mind X XPaperElectronicAction skillOther

Availability of knowledge

Limitations in timeLimitations in spaceLimitations in access X XLimitations in qualityLimitations in form X

Remarks:

22

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Table 3.4: Task Model TM-2 for knowledge of the property industry, neighbourhood amenities, features,etc.

NAME Knowledge of property industry, neighbourhoods, etc.POSSESSED BY Property agentUSED IN 2, 3, 4, 6DOMAIN Real estate

Nature of the Knowledge Bottleneck/to be improved?

Formal, RigorousEmpirical, QuantitativeHeuristic, Rule of thumbHighly specialized, domain-specific XExperience-based XAction-basedIncompleteUncertain, may be incorrect X XQuickly changing X XHard to verify XTacit, hard to transfer

Form of the knowledge

Mind X XPaper XElectronic XAction skillOther

Availability of knowledge

Limitations in timeLimitations in spaceLimitations in access X XLimitations in qualityLimitations in form X X

Remarks:

23

Chapter 4

Organizational Task Analysis

This chapter describes an analysis of the changes in the organization that will result from the proposedsolution. This chapter is organized as follows: Section 4.1 describes the changes to the organization thatwill result from the implementation of the system; Section 4.2 describes the changes to specific tasks;Section 4.3 describes the actions required on behalf of stakeholders; and Section 4.4 describes next steps.

4.1 Impacts and Changes in Organization

The considered knowledge system solution will require changes to the process, resources, and knowledgeof the organization. The structure of the organization, people, and culture and power remain unchanged.

Process

A proposed revised process flow appears in Figure 4.1. The revised process flow includes an accountingfor the knowledge-based system, which will automate the tasks of searching for properties, ranking results,and preparing a report.

24

KM

M:P

RIN

CIPA

L.PR

OP

ER

TY.R

EC

OM

ME

ND

ER.C

OM

(MSc)

Sample

Solution

Explain Recommendations

X

Bill Customer

Assign Request

Elaborate Request

O

E-mail Results

X

X

Print Results

Enter Request OProduce Results and

Rankings Review Rankings Generate Report

Figure 4.1: Proposed process flow, with the new knowledge-based system.

25

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Resources

It is expected that the total process time in the worst case will drop from 75 minutes to approximately 20minutes for corporate clients. For individual clients, the worst-case processing time will be reduced by asimilar amount.

Knowledge

The system will encapsulate the knowledge of how to assign task requests and rank urgency, how to searchweb sites for properties, and how to rank the results.

4.2 Task-specific impacts and changes

The proposed system will require changes in the workflow and dependencies, needed resources, and re-quired knowledge. No other changes are anticipated.

Task layout

Refer to Figure 4.1 for information about the new process flow. The significant changes are in the removalof the search, rank, and report tasks, and replacement with a Produce Rankings and Results task. The taskencapsulates the behaviour of those tasks, and will be performed by the new knowledge-based system.

Resources

The new rankings and results task will require computing resources and may require a centralised serversystem. Staffing resources will be unchanged.

Knowledge

The knowledge required to operate the system will be provided by training sessions. Additional knowledgemay be required on how to maintain the system, or this knowledge could be subcontracted outside thecompany.

4.3 Attitudes and Commitments

The property agents and management would need to buy-in to the system before the project can commence.The agents would need to understand that the system is intended to help them work better, and faster, andprovide a higher quality of service to their existing clients. The system will enable them to spend moretime consulting with clients, and to help them produce a more suitable response as a result. The system hasanother intended effect of increasing revenue, which should please the owner/manager.

26

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

4.4 Proposed Actions

Improvements

The proposed changes to the organization are as follows:

• The organization will need to be trained to use the new system;

• The staff members system will need to be implemented on server hardware;

Accompanying Measures

Training will need to be provided on how to use the system, and on the new workflow. The new system willneed to be advertised to clients if the component to allow clients to input their own requests is implemented.

Expected Results

The following results are expected:

• The processing time for corporate clients should drop by approximately 60%.

• The processing time for individual clients should drop by approximately 55%.

• The processing time for overseas properties should drop by approximately 20%. The drop is less sig-nificant here because the system will not have much expertise in the areas of overseas lets. Additionalmanual processing may be required.

• The quality of the results will be improved.

• The quantity of ranked results will increase.

Managing Change

If business conditions change, the organization should reevaluate the potential for increased business thatresults from the implementation of the system. Because the potential for increased business is part ofthe rationale for implementing this system, a downturn in the market would make delaying this project areasonable course of action.

27

Chapter 5

System Design Document

5.1 Use Case Model

From analysis of the client’s business processes, a use case model was derived that is similar to the to-beprocess model discussed in a previous section. The use case model appears in Figure 5.1. Each use case isdescribed here:

Process Request Actors: Property Agent This use case represents a composition of the workflow items ofsearching, reviewing results, and generating a report.

Enter Request Actors: Property Agent, Client This use case represents the action of entering a new re-quest for service into the system. This action can be performed either by a property agent in responseto a walk-in or telephone inquiry, or by a client through a web-based user interface.

Review Report Actors: Property Agent, Client This use case represents the act of reviewing the reportwith the client, and explaining the results. In some cases the client may wish to review the results onhis own, and ask questions of the property agent later. In that case, the client would execute this usecase.

Review Results Used by: Process Request This use case entails reviewing the results and rankings thathave been generated by the system.

Generate Report Used by: Process Request This use case involves creating a printable or e-mailablereport for the client.

Search Used by: Process Request This use case involves searching the web for appropriate properties. Theuse case is typically an internal action executed by Process Request, but may be manually conductedin some cases.

5.2 Proposed software architecture

The layout of this section is derived from [1].

28

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

System

Client

Property AgentEnter

Request

Process Request

Review Results

<uses>

Generate Report<uses>

Review Report

<<actor>>

Property Web Site

<uses>

Search

Figure 5.1: Use case model for the proposed system.

5.2.1 Overview

The purpose of the system is to realize the use cases described in the previous section. The system ac-complishes this by implementing a web-based user interface for both public and private use, a controller,and a model for retrieving and reasoning over data. The system is capable of formatting results for print ore-mail, and can allow property agents to review and alter result rankings.

5.2.2 Subsystem decomposition

A subsystem decomposition of the proposed system appears in Figure 5.2. The system is decomposed intothe following subsystems:

User Interface This is the user interface for the tool. The UI is a web-based component.

Public UI This is the web-based UI that the clients interact with. It is a specialization of the User Interfacesubsystem.

UI Controller This controls the UI interaction with the model.

Modeller This encapsulates the system’s business logic, and coordinates activities between the knowledgebase and the web retrieval subsystem.

Knowledge base This contains the system’s knowledge, ontology model, information about current prop-erties, and so on.

Information Retrieval This subsystem contains rules for retrieving data from the web site, based on theknowledge model. The subsystem is also responsible for parsing data and formulating appropriatequeries.

5.2.3 Hardware/software mapping

The system is expected to be partially distributed. The system will operate a web-based interface for theclients to enter their own requests remotely, or for the agents to enter requests on behalf of walk-in clients

29

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

System

User Interface

UI Controller

Public UI

Knowledge Base Information

Retrieval

Modeller

<<external>>

Property Web Site

Figure 5.2: Subsystem decomposition for the proposed system.

or phone-in clients. The agent will interact with the web-based interface to produce search results, rankproperties, review the system’s automated rankings, and produce reports. A deployment model of thesystem’s components appears in Figure 5.3.

5.2.4 Persistent data management

The system will maintain persistent data in a combination of relational databases and an OWL ontology.The relational databases will contain the information related to properties that is not captured in the on-tology, or is not relevant to classification tasks. The OWL ontology and related inference systems willallow reasoning over the data as necessary. Because of performance concerns, once a property has beenclassified, it will not be stored in the OWL ontology.

5.2.5 Boundary conditions

This section describes error conditions. Because the system is mission-critical, and must support a highuptime, it is important to discuss possible failure modes of the system. One possible failure is the unavail-ability of the external web sites on which property searches are conducted. In this case, the knowledge-based system can use cached information from prior searches to enable current requests to be serviced.Although the results may not be of a high quality, throughput can be maintained, and additional results canbe retrieved when the web sites return to normal functioning.

In the case where the external Internet connection fails, the system will continue to operate withcached data, as above. Clients will no longer be able to enter their own requests, but the property agents

30

KMM: PRINCIPAL.PROPERTY.RECOMMENDER.COM (MSc) Sample Solution

Agent Workstation

Server <<external>>Client Workstation

Web Client

Knowledge-based System Web browser

Figure 5.3: Deployment of the various subsystems on equipment.

will. In that case, phone-in and walk-in clients will not be affected.If the entire system suffers a catastrophic failure, a fallback to the existing paper-based method

will occur.

31

Chapter 6

References

1. Bruegge, Bernd and Dutoit, Allen H. 2000. Object-oriented software engineering: Conquering

complex and changing systems. Upper Saddle River, NJ: Prentice Hall.

2. Scheiber, G. et. al. 2000. Knowledge Engineering and Management: The CommonKADS Method-

ology. Cambridge, Mass: MIT Press.

32