A Context Sensitive Experience Feeder for Computer Aided

Engineering

Bo Song and Zuhua Jiang

Department of Industrial Engineering and Logistics Management, Shanghai Jiao Tong University, Shanghai, P.R. China

Keywords: Context, Experience, Computer Aided Engineering.

Abstract: Computer aided engineering (CAE) tries to map properties and interactions of real world entities with

symbols and values readable to the machine. Modern CAE software packages are powerful in function, but

users usually need a lot of knowledge and experience to manipulate them. As a kind of tacit knowledge,

experiences require gauged context in order to be fully understood and applied. To better exploit the freely

written, hard-to-encode experiences on the web, we propose in this paper a context sensitive experience

feeding mechanism which is able to recommend experiences matching the context of a given CAE task. Our

method makes use of information extraction and natural language processing techniques to find experience

valuable to engineer’s trouble shooting. Empirical evaluation of a prototypical feeder suggests that our

method is effective.

1 INTRODUCTION

Knowledge employed in so called knowledge-

intensive tasks incorporates two aspects: the explicit

aspect and the tacit aspect. While explicit knowledge

is relatively easy to acquire, codify and reuse, such

operations for tacit knowledge are much harder for

that tacit knowledge is rooted in an individual’s

experience and values and is difficult to reduce to

formal representation (Nonaka and Konno, 1998;

Chen, 2010). Computer aided engineering is a

typical knowledge intensive practice which has

aroused continuous interests among researchers of

knowledge engineering (Colombo, Mosca and

Sartori, 2007). Modern CAE software packages are

powerful in function, but users have to be equipped

with a lot of knowledge and experience to

manipulate them. One proven is that there exist tens

of thousands of questions asking for guidance or

experiences in some online CAE forums such as

XANSYS and iMechanica. In our work, we try to

channel existing experiences to engineers who might

need them in their current working context, hope

doing so could save time that otherwise would be

spent on query formulating, searching, and waiting

for response.

2 CONTEXT SENSITIVE

EXPERIENCE FEED

Knowledge management efforts, to be successful,

need to be sensitive to features of the context of

generation, location, and application of knowledge

(Nidumolu, Subramani and Aldrich, 2001). To make

recommended experience more useful to an engineer,

we should understand what he is doing and what

problem he will face. The information collected

from an ongoing CAE task facilitating such

understanding is called context.

2.1 CAE Task Context

Computer aided engineering maps the property and

interaction of real world entities with symbols and

values readable to machine. To use CAE software

packages to solve engineering problems, people

have to know the terms and concepts denoting such

abstract entities. This makes the name of these

entities a good indicator of what a CAE task is about.

As CAE operations are composed of actions taken

by people on target entity, the verb-object structure

can assume an informative role in describing a CAE

task. Based on these observations, we propose our

context model as following.

343

Song B. and Jiang Z..

A Context Sensitive Experience Feeder for Computer Aided Engineering.

DOI: 10.5220/0004133103430346

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2012), pages 343-346

ISBN: 978-989-8565-30-3

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

The concept list:

C = {c

| i=1, 2, … , m}

where c

denotes a concept (noun phrase)

captured from a CAE task, i=1, 2, … , m

defines the sequence of concepts

The potential trouble pool:

P ={(v , c)| c∈C, v is a verb }

The number of concepts in C is confined to m —

when a new concept is added, the oldest concept in

C is deleted and all subscripts are adjusted. For the

construction of P, other than manually building a

static lexicon, we choose to extract verb-object pairs

from the ever-evolving Internet corpus. To do this,

for noun phrase c

in C, we construct four quoted

queries:

“how to * <c

’s singular form> ”

“how to * <c

’s plural form> ”

“cannot * <c

’s singular form> ”

“cannot * <c

’s plural form> ”

Every query is to be submitted to a search engine

that supports wildcard and exact string match. After

the search engine has returned results for the

constructed queries, we use a part-of-speech tagger

to assign POS tags to matched texts, and any

contained verb-object structures using c

as argument

are extracted according to the following rule (in

BNF):

(“how to”|“cannot”) [adverb]<verb>[pronoun]

[adjective]<noun phrase c

We adopt above query patterns for three reasons:

first, we are mainly concerned with know-how

experiences; second, though there exists other

syntaxes such as “how can/do I/you do a thing” for

people to query know-how knowledge, a few search

trials can tell that with the same semantic meaning,

petitions beginning with “how to” and “I cannot”

surpass others in quantity; third, under the

redundancy surmise of Internet content, a single

syntax should have questioned most aspects of a

concept.

2.2 Trouble Detection

As a CAE task proceeds, the concept list C and

potential trouble pool P are dynamically changing.

At any time the task owner encounters difficulty and

stops to check the experience feeder, we must assess

the task context and come up with remedies for the

trouble that the task owner is most probably facing.

This is done by approximating the probability P(t|C)

for every potential trouble t in P:

(,)

(|)

()

( , )

q max{ ( ) ( , )}

z max{ ( ) { ( , , )}}

ij x

ij i x

wxPt c

wx ev c c





(1)

where

subscript i ranges over the m concepts in

C, and subscript ij denotes the jth verb

that has c

as its direct object ;

≠i ;

w(x) is weight function ;

e(v

) denotes any experience piece

that contains the three keywords: v

, c

and c

;

z is normalizer.

The three approximate equalities each have its

meaning:

 The first approximate equality means the

chance of any specific concept series is treated

as equal.

 The second approximate equality significantly

reduces the scale of joint distribution of

concepts out of computational complexity

concern. Besides, we add a weight function

here to gain some control over the choice of

concepts. An instinctive idea is to use more

recent concepts to infer possible troubles.

 The third approximate equality relaxes the

verb-object constraint between v

and c

when

using them to retrieve evidence. This is

because it is impractical to parse every

sentence while searching a gigantic corpus.

Whenever a new concept is captured and

changes C, we recalculate P(t|C) for every potential

problem in P. For some most probable troubles, we

would search for relevant experiences and

recommend them to the task owner.

2.3 Experience Retrieval

To retrieve a piece of experience as candidate

remedy for sensed trouble (suppose the most

troublesome is t

), we use three features of textual

experience for match making: trouble mentioning,

context overlap, and procedural marks.

Trouble mentioning: an empirical remedy should

explicitly mention the target trouble, t

, which is

described by a verb-object pair. To do this, a natural

language parser is needed.

Context overlap: for a piece of textual experience,

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

344

Figure 1: Context sensitive experience feeding.

e, we count the times that each concept in C appears

in e and denote it with f

i=1, 2, … , m

. Then context

overlap index for e can be calculated by:

= ∑

)

(2)

where L

denotes the words count of e

Procedural marks: since useful experience usually

takes the form of procedural guidance, we studied

literature investigating the characteristics of

procedural text and use the linguistic marks

proposed in (Aouladomar, 2005) and (Fontan and

Saint-Dizier, 2008) to assess the quality of a

candidate experience.

Retrieved experience pieces are firstly grouped

based on what trouble they address. These groups

are then ranked in descending order according to the

product of trouble probability and averaged context

overlap. Experience-carrying texts within each

group are sorted descendently according to the

number of procedural marks they contain. The

information process flow for experience feeding is

shown in Figure 1.

3 METHOD EVALUATION

3.1 Method Implementation

In our research, we choose ANSYS based finite

element analysis as an exemplary CAE task.

ANSYS is a wildly used computer program for

doing finite element analysis, and it is operated in

such way that a series of text-rich dialog windows

are gone through one by one as the task stage

advances. Thus an agent able to record texts on

ANSYS dialog window can be used for context

collecting. The whole procedure of experience

feeding is shown in Figure 1.

Aside from the description presented in section 2,

we use following concrete configurations when

realizing an ANSYS experience feeder:

 m, the number of memorized concepts, equals

12 ;

 recorded window titles are assigned POS tags

for noun phrase extraction;

 for each noun serving as context, we retain the

4 most appeared verb (counted from each

query’s first Google 100 results);

 when using formula (1) to find the most

probable potential problem, e(v

) is any

Google result in which the three keywords co-

occur, and the weight function is set as below:











1/2

0 

(3)

 for a perceived trouble, t

, relevant experience

is picked from the Google results used to

assess the probability P(t

|C) — if parsed

e(v

) gets the trouble mentioning feature,

then the paragraph it resides in is treated as

one relevant experience ;

 Stanford Parser (Klein and Manning, 2003) is

used to parse Google results and assign POS

tags.

3.2 Empirical Evaluation

Totally 24 college students in their junior year

taking a finite element analysis course were

materially rewarded to assess the effectiveness of

experience feeder. We chose a static contact

problem which requires about 20 steps to complete

as background task, and recorded the ANSYS

window titles each student had gone through when

solving the task. Each time an ANSYS window is

activated, the recommendation list is refreshed and

the 10 most highly ranked experiences are shown.

Table 1 is a snapshot of experiences recommended

when the 7

window is activated.

Each student is asked to read through the

experience lists and rate each experience with three

grades: 1) useful to current problem solving; 2) good

to know but not directly useful to current task; 3)

totally irrelevant. The evaluation result is shown

below (proportion=0.2 means ).

Number of instantly useful ratings among 24 participants

rating count 0 1 2 3 4 5 >5

person count 1 5 6 6 4 1 1

Proportion of relevant ratings among 24 participants

proportion 0.1 0.2 0.3 0.4 0.5 0.6 0.7

person count 1 3 6 6 4 2 2

AContextSensitiveExperienceFeederforComputerAidedEngineering

345

Table 1: Recommended experiences.

Window title history Relevant experiences

Global Element Sizes

For trouble: generate mesh concerning: area

Typically, you will generate a mesh for the source area yourself, before you

sweep the volume.

For trouble: generate element concerning: contact

1. Sets of nodes that are likely to come into contact must be defined and used

to generate the necessary elements.

2. Generate contact elements whose contact node is within a radius of RADC

measured from the centroid of each target element face.

For trouble: find area concerning: mesh

While meshing the model in ANSYS we are facing difficulty in finding

missing common areas.

…

Meshing Attributes

Volume Sweepings

Mesh Volumes

Contact Manager

Contact Wizard

Select Areas for Target

4 CONCLUSIONS

Experiences are not formal, validated knowledge,

but they are highly thought-provoking and worth

sharing. In this paper we have proposed an

experience feeder for engineers who process finite

element analysis, a typical CAE task. Experiment

shows that by our method engineers can gain

knowledge about what problems they may encounter

at different task stage and what the possible

solutions are. Other study has achieved higher

performance in knowledge recommending (Shen,

Geyer, Muller, Dugan, Brownholtz and Millen,

2008), but they require training on manually

annotated activities and resources, which we do not.

In future work, to enhance our method, geometry

feature recognition can be used to capture more

informative concepts form CAE tasks, and keyword

extraction technique can be used for extracting

concepts from task defining documents.

ACKNOWLEDGEMENTS

The author is most grateful to National Nature

Science Foundation of China (No. 70971085) and

the Research Fund for the Doctoral Program of

Higher Education of China (No. 20100073110035),

for financial support that made this research possible.

REFERENCES

Aouladomar, F. (2005). Towards answering procedural

questions. In Knowledge and Reasoning for Answering

Questions, Workshop associated with IJCAI05, 21-31.

Chen, Y. (2010). Development of a method for ontology-

based empirical knowledge representation and

reasoning. Decision Support Systems,50, 1–20.

Colombo, G., Mosca, A. and Sartori, F. (2007). Towards

the design of intelligent CAD systems: an ontological

approach. Advanced Engineering Informatics, 21,

153–168.

Fontan, L. and Saint-Dizier, P. (2008). Constructing a

know-how repository of advices and warnings from

procedural texts. In Doceng'08: Proceedings of the

Eighth Acm Symposium on Document Engineering,

249-252.

Klein, D. and Manning, C. D. (2003). Accurate

unlexicalized parsing. In Proceedings of the 41st

Meetingof the Association for Computational

Linguistics, 423-430.

Nidumolu, S. R., Subramani, M. and Aldrich, A. (2001).

Situated learning and the situated knowledge web:

exploring the ground beneath knowledge management.

Journal of Management Information Systems, 18(1),

115-151.

Nonaka, I. and Konno, N. (1998). The concept of ‘ba’:

building a foundation for knowledge creation.

California Management Review, 40( 3), 40-54.

Shen, J., Geyer, W., Muller, M., Dugan, C., Brownholtz,

B. and Millen, D. (2008) Automatically finding and

recommending resources to support knowledge

workers' activities, In ACM 2008, 207-216.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

346