Hybrid System for Collaborative Knowledge Traceability
An Application to Business Emails
Francois Rauscher, Nada Matta and Hassan Atifi
Institute ICD/Tech-CICO, Université de Technologie de Troyes,
12 rue Marie Curie, BP. 2060, 10010, Troyes Cedex, France
Keywords: Knowledge Management, Traceability, Problem Solving.
Abstract: In this paper we propose a Knowledge Traces Retrieval (KTR) system. It addresses the problem of retrieving
elements of problem solving and design rationale inside business emails from a project. Even if knowledge
management tools and practises are well spread in industry, they are rarely used for small projects. Our system
aims at helping user retrieve traces of problem solving knowledge in large corpus of email from a past project.
The framework and methodology is based on enhanced context (project data, user competencies and profiles),
and use machine learning technics and ranking algorithm.
1 INTRODUCTION
Knowledge management became very popular in
organizations in the 2000s after the works and
discoveries of the past decades (Steels, 1993; Van
Heijst et al., 1998). With the growth of the Internet in
the 1990s, the dissemination and sharing of
information has increased greatly. Many firms have
the ambition to be able to apply this to knowledge.
Companies have actually recognized the strategic
value of knowledge as intangible capital (Grundstein,
1995) that could lead to competitive advantages.
However the path from Information management
to Knowledge management is complex and involve
the developments of systems that allow the firms to
recognize, create, transform and distribute
knowledge. The goal of these corporate memories is
always the same: in one of the company's operations
expertise has been created and/or used. This
knowledge should not be lost because it can be used
in similar areas for the company, and sometimes in
different ones (by analogy or conceptualization), or in
terms of traceability to determine by whom and why
a decision was taken and what procedure has resulted.
Information technology is an essential element to
mobilize social capital for creation of knowledge, but
how to proceed when the people involved in a project
have left the company? This is especially true in
software design projects. Software engineering is a
quickly evolving, knowledge-intensive business
involving many people working in different phases
and activities (Rus et al, 2002). In this study, we
focused on companies with software design and
development projects involving geographically
distributed teams, or remote workers. These
companies have consulted us with identified needs in
terms of traceability and knowledge reuse but with
limited resources. The projects were completed
(sometimes years ago), there were the
deliverables/products, project management data
(planning, documentation, and specifications), the list
of participants and their skills, and their overall
electronic exchange with related documents. Note
that no recording (audio or video) of meetings or
telephone conversations and videoconferencing were
available.
A typical situation occurring was a manager
asking to an employee: “What is the exact file format
in this use case of the software XYZ and why?” And
the only solution for the employee is to explore tens
of gigabytes of emails requesting by keywords to try
to find useful information. We present here a system
called KTR that aims at helping users retrieving
relevant (according to a query) traces of problem
solving knowledge among a large emails corpus.
Analysis method described in our previous work
(Rauscher et al, 2002) suggested that taking into
account the enhanced user context, organization,
roles... can be used to find usable traces of problem
solving knowledge.
The rest of this paper is organized as follows. In
Section 2 related work is discussed briefly. Section 3
260
Rauscher, F., Matta, N. and Atifi, H..
Hybrid System for Collaborative Knowledge Traceability - An Application to Business Emails.
In Proceedings of the 7th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2015) - Volume 3: KMIS, pages 260-267
ISBN: 978-989-758-158-8
Copyright
c
 2015 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

introduces our proposal then explains our approach
and choices. Section 4 describes our system
architecture and algorithm. A real life application is
given in Section 5 and Section 6 comprehends our
conclusions.
2 RELATED WORKS
2.1 Emails and Tasks
Emails is closer to written language (Baron, 1998),
but in a computer mediated communication approach,
as stated by Herring, et al. (2004), it can be considered
as a good candidate for using discourse analysis
techniques.
Collaborative Problem solving often leads to
people assigning tasks to others. A study from Kalia,
et al. (2013) on Enron corpus presents a method to
identify and track tasks and commitments inside
business emails. This was done by pure NLP technics
like using n-grams, part of speech tagging and
machine learning. Our system enhanced the context
with user competencies and roles, project
organization and phases. One can also notes the work
of Scerri, et al. (2010) that propose a system called
Semanta to assist users during their daily emails
workflow to track actions items like Meeting
Request, Task Assignment, and File Request.
Traceability of requirements in software
development is usually done through the usage of a
simple matrix but does not keep the record of the
“how, why and who” during the design and
implementation process. As our approach uses all the
aspects of the project, phase, roles and not only
textual content, it is closely related to traceability in
the perspective of project memory.
2.2 Project Memory
Compare to corporate memory, project memory (PM)
is a restricted part of a much larger capitalization
exercise of a whole range of diverse experiences
within the company. A project memory is generally
defined as a representation of the experience gained
during the implementation of projects (Dieng, et al.,
1998). It describes the history of a project and the
lessons learned during the lifetime of a project
(Pomian, 1996). This memory must contain elements
of experience from both the context of problem
solving and its resolution. Project memory aims at
traceability and reuse in similar project. Project
memory used as materials the project data, the
products of the project, stakeholders (roles,
competences in the organization, exchanges and
meetings, electronic communications, telephone) and
related documents. However a reverse scale effect
might appear compared to corporate memory: it is
sometimes difficult to establish a method, software
and collection of interviews for a project involving
fewer than a dozen people. Tools and procedures exist
for such tasks, but the size of the teams makes the
systematic collection expensive and difficult.
Especially when the project is finished and the team
dismissed. In (Matta, et al., 2010) linguistic
pragmatics was used on discussion forums to identify
criteria that help analyzing messages of coordination
in design project. KTR system will use context
knowledge in a similar way.
3 TRACES OF KNOWLEDGE
As we stated in the introduction, the goal of our KTR
system is to help the user retrieve “traces of
knowledge”. We define Knowledge Traces (KT), as
messages containing meaningful information
regarding the team’s members having a problem-
solving mediated exchange over email.
As a typical use, the user will input a query and
the KTR system will present a list of messages from
the email corpus matching the query and having a
high score of being part of collaborative problem
solving between the project’s members. Retrieval
process can be viewed as ranking or classification
problem, in this study we consider it as ranking
problem. We will compute a score on each message
based on the user query and the KT elements. In order
to decide if a message contains KT, we will check if:
 The message is dealing with topics from the
project
 The message thread contains at least a request
(problem statement)
 The messages in the same thread following the
initial request contains elements of answer or a
decision.
3.1 Topics
We called topics a lexicon of words regarding the
project. This lexicon can be built from the following
sources:
 Project phasing and specifications documents ;
 domain ontology if available
 an expert;
For instance we could decide that the lexicon will
contains the topic XML (because it was an important
Hybrid System for Collaborative Knowledge Traceability - An Application to Business Emails
261

milestone in the project) defined by a list of keywords
(e.g: xml, tag, tree, xsd, dtd, schema, markup,
structuration ...). When we will match any of the
keywords in a message, we will know the message
deals with the topic XML. The name of the topics in
themselves are not relevant, there are simply a
convenient way of handling them.
3.2 Problem Solving: The Request
As stated above, we chose first to focus on problem
solving (Newell, et al., 1972) because useful
knowledge for the business is most likely to be used
or created during this kinds of exchanges. Especially
on the undefined types of problem or "wicked
problems" (Shum, 1997; Conklin and Weil, 1997),
where collaborative knowledge occurs naturally.
Hardin (2002) distinguish three main components in
problem-solving: givens (facts and context), goal and
operations (action to be performed).
In software development, when a team is trying to
fulfill the specifications and requirements of the
product or to correct a bug, it involves abstract and
cognitive tasks. The first one being by fully
qualifying the demand (the “goal”), then stating the
problem to reach it. The “operations” will occur in the
following exchanges when the team is going to face
and solve the problem (see section 3.3 on the
solutions). As the teams are using computer mediated
communication, we have to carefully examine the
requests in the email threads.
Finding requests requires interpretation of the
intent that lies behind the language used. We chose to
approach the problem as one of speech act detection.
The theory of speech acts stems from the original
works in philosophy of language from Austin (1975)
and Searle (1969). Since then many studies have been
conducted on speech acts (and particularly request),
in different disciplines such as theoretical linguistics
and natural language processing (NLP) with for
instance works on automated speech acts
identification in emails (Carvalho, et al., 2005;
Lampert, et al.,2010). De Felice, et al. (2012) noted
there is very little concern with data other than spoken
language.
In the present study we narrowed our research to
the analysis of the act of requesting in problem
solving sequences. In pragmatic linguistic, request is
part of the directive illocutionary acts. An
illocutionary act being an act that is performed by
saying it. The purpose of a request is to “get the hearer
to do something in circumstances in which it is not
obvious that he/she will perform the action in the
normal course of events” (Searle, 1969). At first level
we can we can distinguish between the request for
saying (e.g. “could you me tell me”) and the request
for a doing (“Could you do that”). These types of
request are quite common among professional email
exchanges.
However illocutionary acts have different
enunciative modalities: There is not a one to one
relationship between a speech act and its linguistic
realization. As we show above a request can be
performed linguistically in several ways: "do that",
"could you do that?", and «I would like you to do
that". A direct request may be imperative (order) or
performative (like an assessment of obligation or
need). An indirect request may question the capacity,
the willingness, etc... of the hearer or give a
suggestion. Often indirect requests are used in
professional context because request in its inner
nature is an FTA (Face Threatening Act) (Brown, et
al., 1978; Goffman, 1959) that can be soften by
politeness.
Our approach following resolutely a discourse
analysis, we are working on statements that can be
properly interpreted by taking into account both the
discursive context (i.e. what constitutes the content of
the email message) and context of “utterance” (that is
to say, the situation of dialogue). The linguistic
markers alone are not sufficient to indicate what is
actually done in a computer mediated communication
situation. As a side remark, it could be interesting in
future studies to look at other types of speech acts
(e.g. promissive, assertive...)
3.3 Solutions
When the system have detected a potential request
regarding the projects topics, we would like to track
the exchange between the team members to keep a
trace of arguments, related matters, decisions and
possible solutions to the initial problem. Our simple
hypothesis is that collaborative knowledge is more
likely to be created when some of people exchanging
messages have the necessary competencies to solve
the current problems.
Competency definition depends highly of the
discipline (sociology, psychology, management) as
stated in (Harzallah, et al., 2002; Vergnaud, et al.,
2004). In the perspective of human resources,
competencies are the measurable or observable
knowledge, skills, abilities, and behaviors (KSAB)
necessary to achieve job performance. One can
distinguish between soft competencies (managerial
and social interaction), hard competencies (functional
and technical specific to a field (Tripathi and
Agrawal, 2014). In our model, we will focus on
KMIS 2015 - 7th International Conference on Knowledge Management and Information Sharing
262

technical competencies and their relations to the tasks
that must be accomplished for the project.
4 KT-RETRIEVAL SYSTEM
In this section, we will go into further details
explaining which kind of information is used and how
the ranking for retrieval is calculated.
4.1 Overview
KTR follow a two-step approach, first indexing the
messages, then ranking the relevant ones according to
a user query. In order to do the indexing step, we first
compute for each message a feature vector composed
of a topic part, a request part, and a solution part. This
gave us a KT score for each message. For the ranking
part we use a linear combination of the KT score with
the basic matching score between the user query and
the message.
The overview of KTR system is illustrated in
Figure 1.
4.2 Message Feature Vector
An email corpus is a set of messages ordered by time
of arrival and grouped by threads (i.e. initial message
with replies).
In a thread T, consisting of messages (Mi)
i

T
,
each message M as an emitter E
M
and receivers
R
M
=(TO
M
,CC
M
) (respectively direct receivers (TO)
and carbon copy (CC)).
As in Carvalho and Cohen (2006) we preprocess
the messages to make them suitable for parsing and
extracting features. For message in the same thread
we remove the duplicate part in case of reply or
forward (e.g. quoted reply content) and the signature
or disclaimer part. The remaining parts are: the
subject of the email, the sanitized body, the name of
attachments.
4.2.1 Topics Part
Using existing context knowledge (project phases and
specifications), we build the topics lexicon. This
lexicon is voluntarily kept simple and have the form
L = (t
i
)
0≤i<t
, where t is the number of topics:
Topic t
i
: keyword
i1
, keyword
i2
... keyword
ip
.
Keywords chosen in topics shall not overlap too much
to keep the results significant.
The content of all messages are represented by
Vector Space Model (Salton and Buckley, 1988), i.e.
M= (w
i
)
0≤i<k
, in which each term is weighted by its
tfidf (term frequency, inverse document frequency)
score, k being the size of the vocabulary. The same is
done for topics. We then compute a ranking between
our messages and each topics using a cosine
similarity based algorithm. This give us a topics
matrix T where (T
ij
) represents weight of topic j in
message i. The score of topic part for a message M
i
will be:
Topic_part

M
i


1
t


1
j
t
(1)
4.2.2 Request Part
The request detection is a well-known non trivial
problem. We took a simple approach similar to
Lampert, et al. (2010). However we worked at
sentence level and if a request was detected in any of
the sentences, the message was classified as request.
Figure 1: Overview of the KTR system and structure.
Hybrid System for Collaborative Knowledge Traceability - An Application to Business Emails
263

Sentence splitting was done according to punctuation
and paragraph signs.
We chose customs parameters and an SVM
classifier, implemented using Azure ML (Microsoft
cloud machine learning platform). Given sentences of
an email message as input, our binary request
classifier predicts the presence or absence of request
utterances within the sentence.
In order to do so, we establish with a linguist a set
of custom features for each sentence. Presence or
absence of: interrogation sign, specific bigrams and
trigrams patterns based on pragmatic (“you should”,
“we must”, “can you”, etc.) and keywords
(“question”, “problem”, “error”, “who”, “what”,
“how”, etc.). We also add a temporal part by taking
into account if a request sign was already present in
previous messages from the same thread. When users
are facing a problem, they are usually asking a lot of
questions before reaching an agreement on a solution.
Another axis of analysis are the relationships of
the project members. Roles in the organization are
important to our study because they could help detect
indirect requests. For instance, if a manager is writing
to a developer “I would like (...)”, it is for us the sign
of an implicit request. Our model takes the roles into
account using official function (hierarchical) or
business relations (client/contractors). We built a
“relationship” matrix R by using a weighted directed
graph representing both hierarchical and
client/contractors links. Users are vertices of the
graph and the weights on edges bring a measure of
user/user “influence” (real values between 0 and 1).
R
ij
stands for the “ordering capacity” from user i to
user j. We use the max over to TO receivers.
The features we use in our SVM request classifier
for a sentence are:
 Presence of verbal signs patterns (e.g., might,
may, should, would, do you, etc..
 Presence of specific keywords
 Presence of interrogation mark
 Presence of previous request in same thread
 Influence score of emitter/receivers
It is important to note that this classifier is not project
specific and once trained can be use in other projects.
Finally the score of the topic part is computed as
in Equation (2), 1 if a request sentence was found in
message and topic_part >0 (we discard requests not
related to the project’s topics), 0.5 if the message was
part of a thread where a previous request was found,
0 elsewhere.
Request_part

M
i



1ifrequesttopic
0.5postrequest
0notrequest
(2)
4.2.3 Solution Part
We look for pieces of knowledge related to the
requests founds in the previous step. For each thread
T, we have identify messages M
r
where a request is
likely to occur. We will then examine all the
following messages in the same thread i.e. M
s, T
=
(M
i
)
(i>r), T
. taking into account user competencies.
First we built a matrix CU representing user
competencies, (using curriculum vitae and function
description of their role in the project). CU = (CU
ij
)
representing the skill level of user i in competence j.
We took similar approach as Vergnaud et al., (2004)
to measures skills (0=not knowing, 0.25 =novice,
0.5=medium, 0.75=experienced, 1= expert).
Then we built a matrix CT representing the
competencies needed to fulfill a topic (its associated
tasks) for the project, CT = (CT
ij)
representing the
importance of competency i regarding the topics j.
This matrix is built with experts in each topics, again
with discrete weighting (ranging from 0=competency
useless for the tasks, 0.25, 0.5, 0.75, 1= vital
competency). We then construct the matrix UT =
t
CU.CT where (UT
ij
) stands for a very rough
estimation of the skills of user i regarding the topic j.
The matrix UT is normalized using Frobenius
norm not to depend on number of competencies or
users. We compute on each message M
i
 M
s, T
, its
solution score:
Solution_part

M
i


1
t
UT
ij
.T
ij

1jt
(3)
We are dealing messages with potential solution
of the problem solving raised by request in M
r
, we are
trying to assert that the emitter have the necessary
competencies to bring new knowledge regarding the
current topics.
4.3 KT Score and Ranking
The KT Score is calculated using the previous sub
scores on each message.
KT_Score

Mi



Topic_partM
i

Request_partM
i

Solution_partM
i

(4)
This score evaluates the relevance of message of
being part of a problem solving trace in the project.
In order to do the final ranking for the retrieval
step based on user query Q, we have to use also the
basic cosine similarity being sim

,Q (in Vector
Space Model with tfidf weighting) between user
query and message 

. This is to take into account the
specific terms of the user query
KMIS 2015 - 7th International Conference on Knowledge Management and Information Sharing
264

Finally the global ranking r for a message 

is a
linear combination which is calculated as follow:





μ_



1‐μsim

,Q
(5)
Where μoμ1is a combing parameter. If
μ=0, we reverse back to simple tfdidf similarity
ranking like Lucene text-search engine (Gospodnetic,
2004)
4.4 Algorithm
The overall algorithm for indexing and ranking is
described in Figure 2. The global ranking is giving a
relevance score for each message. We output the
messages in decreasing ranking order but for better
understanding for the user we kept them grouped by
threads of conversation.
Figure 2: KTR Algorithm.
5 EXPERIMENTS
A publishing group editing magazines and books had
in 2009 a software design project that lasted 2 years.
A remote software development company was hired
to create a workflow tool for journalists and lawyers.
Nearly all the communications during specification,
implementation, tests and delivery were done through
email. The corpus was collected 2 years after the end
of the project.
5.1 Corpus and Project
The corpus represent 3080 messages/ 14987
sentences in 801 threads between 30 projects actors
The team was split between the contractors and the
development team. Among these, various roles and
skills were present, but the main actors were:
 U1: Chief editing manager (skill: law and
management, Role: Contractor);
 U2: Law Journalist (skill: law and management,
Role: Contractor end-user);
 U3: Information System Manager (Skill:
Information system, Role: Contractor);
 U4: Information System Project Manager (Skill:
Information system, Role: Contractor Employee);
 U5: Information System Developer (Skill:
Software Engineering, Role: Development
manager).
5.2 Application Settings
A topic lexicon was built according to section 4.2.1
and the message topic T matrix was computed
accordingly. We had 10 topics. Stemming was done
to compute the term frequency and cosine similarity.
For request, we built a feature vector as described in
section 4.2.2 for each sentence containing (pragmatic
verb markers, specific keywords, interrogation mark,
temporal part, max emitter influence over TO (direct)
receivers). For the ranking algorithm, to compute the
global ranking combination parameter μ was set to
0.4. In future works, we will evaluate the impact of
this parameter. Finally the two matrices CU and CT
defined in section 4.2.3 were computed. This
operation is done once and valid for the whole
project.
On Table 2 and 3, we can see for instance that the
user U1 possess good competencies in law and some
in InDesign, and that these competencies are
important for the tasks in topics Code and Paper.
Table 1: Excerpt from topic lexicon.
Topic Keywords
Code
Law, legifrance, insurance, chapter, article,
annexes, labor code
Paper Indesign, Xpress, print, template, styles, margin
Inputs: Project Data, Corpus, User competency
Output: KT_Score for message
1 Indexing:
2 Prepare Topic lexicon L from Project Data
2 Prepare Topic matrix T (section 4.2.1)
3-Prepare CU,CT and UT matrix (section 4.2.3)
4-Prepare Influence Role R matrix (section 4.2.2)
5-Train SVM for request sentences
6 For each thread T  Corpus
7 For Each message MT
8 Compute Topic_part(M) (Equation (1))
9 Compute Request_part(M) (Equation (2)
10 Compute Solution_part(M) (Equation (3)
11 Compute KT_score(M) (Equation (4)
12 end
13 end
14 Retrieving:
Inputs: User Query Q, KT_Score, Corpus
Output: Global ranking and messages
15 For Each message MCorpus
16 Compute sim(Q,M) (section 4.3)
17 Compute rank(M) (Equation (5))
18 end
19 Output messages by decreasing rank grouped by
thread.
Hybrid System for Collaborative Knowledge Traceability - An Application to Business Emails
265

Table 2: Excerpt Competency User matrix.
Table 3: Excerpt Competency Topics matrix CT.
5.3 Corpus Labelling
In order to evaluate KTR system, we had to label the
corpus. At first this was done by a linguist looking for
the presence of direct/indirect requests.
Then we had to label the messages where answers to
these requests were explored during the problem
solving. In that case the ground truth or gold standard
is estimated from the subjective opinion of a small
number of experts. At first a technical expert was
asked to label messages that correspond to
collaborative knowledge traces, then a manager that
worked on this precise project 4 years ago did the
same. The manual labelling is an expensive and time-
consuming process. All the messages selected by both
experts were considered as good candidate for
containing solutions to problems.
This part was especially tedious and due to time
constraint only 70% of the corpus (starting from
beginning in emails date order) was annotated by
these last two experts.
5.4 Evaluations and Results
To measure the performances, we decided to compare
KTR with baseline cosine similarity (Lucene-like text
search). This was done by setting µ parameter to zero.
We forged 40 queries based on real life scenario (for
instance “xml file format insurance law comment”)
and compared the results of the two methods by
counting the numbers of results labelled by experts in
the 20 first answers. The dimension of the feature
vector was 25. We found 7% improvement over the
keywords only search on theses queries. We also
noticed that request part is bringing noise (giving high
relevance to messages containing request but not
always related with the keywords of the query) and
fine tuning would be necessary.
6 CONCLUSION
In this article we proposed KTR system for analysing
professional emails from a project in order to help
users locate problem-solving traces. Our study aims
at traceability in the context of project memory. In
contrast to previous models of extracting meaningful
elements (like tasks, or concepts, etc...) from emails
by using pure NLP techniques, we emphasize on
enhancing the context and incorporate pragmatics and
organizational components (speech acts, competen-
cies modelling, roles).
Numerical results shows that the task of detecting
problem solving knowledge traces is very delicate
and subjective. However this task is necessary and is
accomplished by employee in everyday life. This
results further indicate that interesting pieces of
knowledge does not always came from an initial
request, and that requests and answers are often
interweaved creating noisy context. Still there are
some issues regarding the granularity (sentence or
message level) of the approach requiring future study
to achieve a better understanding of patterns of
exchanges during collaborative knowledge creation.
While this study has shown relatively interesting
results, we are planning further research studies to
investigate the possibility of improving the
algorithmic part. First by enlarging the context to
surrounding messages and relevant threads and taking
into account attachment’s content. Second by making
a more precise match between topics of request and
possible answers. It has become clear that taking only
into account the textual content of email (and not the
overall human context surrounding it) would lead to
very limited results as detecting traces of
collaborative knowledge. The proposed system will
also be evaluated on scenario where the project is yet
not completed and will involve the users. Ultimately,
our work contributes to project memory: traceability
and structuration of knowledge in daily work
realization of project.
U1 U2 U3 U4 U5 U6 U7
XML/XSL 0 0 0 0 1 1 0
C# 0000100
SQL 000,250100
Architecture0000100
LawCode 1 0,75 0 0 0 0 0,5
LawWriting 1 1 0 0 0 0 0
Indesign 0,25 0 0 0,25 0,5 0 1
HTML 0 0 0 0,25 1 0,25 0
XML BDD Workfl o
w
Code  Paper
XML/XSL 1 0 0 0 0,5
C# 0,25 0 1 0 1
SQL 0 1 0,5 0 0
Architecture 0,5 0,5 0,5 0 0
Lawcode 0 0,25 0,5 1 0
Lawwriting 0 0 0 1 0
Indesign 0 0 0 0 1
HTM L 0 0 0 ,5 0 0
KMIS 2015 - 7th International Conference on Knowledge Management and Information Sharing
266

REFERENCES
Austin, J. L., 1975. How to do things with words (Vol. 367).
Oxford university press.
Baron, N. S., 1998. Letters by phone or speech by other
means: The linguistics of email. In Language &
Communication, 18(2), pp. 133-170.
Brown, P., Levinson, S.C., 1978. Universals in language
usage: politeness phenomena. Question and politeness,
ed. by Esther N. Goody. Cambridge: Cambridge
University Press.
Carvalho, V. R., and Cohen, W. W., 2005. On the collective
classification of email speech acts. In Proceedings of
the 28th annual international ACM SIGIR conference
on Research and development in information retrieval
pp. 345-352.
Carvalho, V. R.., and Cohen, W. W., 2006. Improving
email speech acts analysis via n-gram selection. In
Proceedings of the HLT-NAACL 2006 Workshop on
Analyzing Conversations in Text and Speech (pp. 35-
41). Association for Computational Linguistics.
Conklin, E. J., & Weil, W., 1997. Wicked Problems:
Naming the pain in organizations (White Paper): Group
Decision Support Systems.
De Felice, R., Deane, P., 2012. Identifying speech acts in
emails: Toward automated scoring of the TOEIC®
email task. ETS Research Report No. RR-12-16,
Princeton.
Dieng, R., Giboin, A., Amergé, C., Corby, O., Després, S.,
Alpay, L., and Lapalut, S, 1998. Building of a corporate
memory for traffic-accident analysis. In AI magazine,
19(4), p.81.
Goffman, E. 1959. The presentation of self in everyday life.
Gospodnetic, O., and Hatcher, E., 2004. Lucene in Action.
Manning Publications.
Grundstein, M., 1995. La capitalisation des connaissances
de l‘entreprise, systèmes de production des
connaissances. In Acte du colloque de l ‘Entreprise
apprenante et les Sciences de la Complexité, Aix-en-
Provence.
Hardin, L. E., 2002. Problem Solving Concepts and
Theories. In Journal of Veterinary Medical Education,
30(3), pp. 227-230.
Harzallah, M., Leclère, M., & Trichet, F., 2002.
CommOnCV: modelling the competencies underlying
a curriculum vitae. In Proceedings of the 14th
international conference on Software engineering and
knowledge engineering pp. 65-71.
Herring, S. C., Barab, S., Kling, R., & Gray, J., 2004. An
Approach to Researching Online Behavior. Designing
for virtual communities in the service of learning, p.
338.
Kalia, A., Motahari Nezhad, H. R., Bartolini, C., & Singh,
M., 2013. Identifying business tasks and commitments
from email and chat conversations. In tech. report, HP
Labs.
Lampert, A., Dale, R., and Paris, C. 2010. Detecting emails
containing requests for action. In Human Language
Technologies: The 2010 Annual Conference of the
North American Chapter of the Association for
Computational Linguistics (pp. 984-992). Association
for Computational Linguistics.
Matta N., Ermine, J-L., Aubertin, G., Trivin, J., 2002.
Knowledge Capitalization with a knowledge
engineering approach: the MASK method. Knowledge
Management and Organizational Memories, Dieng-
Kuntz R., Matta N. (Eds.), Kluwer Academic
Publishers.
Matta, N.; Atifi, H.; Sediri, M.; Sagdal, M., 2010. "Analysis
of Interactions on Coordination for Design Projects," In
Sixth International Conference on Signal-Image
Technology and Internet-Based Systems (SITIS),
pp.344-347.
Newell, A., Simon, H. A., 1972. Human Problem Solving.
New Jersey: Prentice-Hall, Inc.
Pomian, F., 1996. Mémoire d’entreprise, techniques et
outils de la gestion du savoir. Sapientia,
Rauscher F., Matta N., Atifi H., 2014. Discovering
Problem-Solving Knowledge in Business Emails -
Traceability in Software Design Using Computer
Mediated Communication in Proceedings of KMIS
2014/IC3K.
Rus, I., & Lindvall, M., 2002. Guest editors' introduction:
Knowledge management in software engineering.
IEEE software, (3), pp. 26-38.
Salton, G., and Buckley, C., 1988. Term-weighting
approaches in Automatic text retrieval. Information
processing & management, 24(5), pp. 513-523.
Searle, J. R., 1969. Speech acts: An essay in the philosophy
of language (Vol. 626). Cambridge university press.
Scerri, S., Gossen, G., Davis B., and Handschuh S.; 2010.
Classifying action items for semantic email. In
Proceedings of the 7th Conference on International
Language Resources and Evaluation (LREC).
Shum, S. B., 1997. Representing hard-to-formalise,
contextualised, multidisciplinary, organisational
knowledge. In Proceedings of the AAAI Spring
Symposium on Artificial Intelligence in Knowledge
Management.
Steels, L., 1993. Corporate knowledge management. In
Proceedings of ISMICK 1993, Compiegne.
Tripathi K. and Agrawal M., 2014. Competency Based
Management. In Organizational Context: A Literature
Review. In Global Journal of Finance and
Management. ISSN 0975-6477 Volume 6, Number 4
pp. 349-356.
Van Heijst, G., van der Spek, R., and Kruizinga, E., 1998.
The lessons learned cycle. In Information technology
for knowledge management. Springer Berlin
Heidelberg. pp. 17-34.
Vergnaud, N., Harzallah, M., and Briand, H, 2004. Modèle
de gestion intégrée des compétences et connaissances.
EGC pp. 159-170.
Hybrid System for Collaborative Knowledge Traceability - An Application to Business Emails
267