IFAC-PapersOnLine 49-12 (2016) 473–478 ScienceDirect ScienceDirect Available online at www.sciencedirect.com 2405-8963 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2016.07.658 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Keywords: Knowledge Management, Traceability, Problem Solving, e-mails, project memory. 1 INTRODUCTION Computer mediated communication is ubiquitous in Software design projects. Email is used for project coordination, but also for design, implementation and test. Especially with currents agile development methods, it is very common to interact through computer meditated communication like email, instant messaging and other collaborative tools in order to express functional needs, notify issues and take appropriate decisions. The knowledge produced in these types of interactions is often buried inside email boxes, hence being volatile and not easily reused. Project Memory (Matta et al, 2000) aims at describing organizational and cooperative dimensions of knowledge created during the lifecycle of a project. Our objective is to discover if we can structure and extract knowledge from professional emails in order to trace some phases and decisions inside a project memory. The main questions of our research are: Is it possible to extract knowledge from e-mails? Which techniques can be used for that? Are current knowledge engineering techniques adequate for this type of situations? For instance: a manager asking to an employee: “What was the exact file format adopted in project “a” 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. To answer this type of question, we develop a system called KTR that aims at helping users retrieving relevant (according to a query) traces of problem solving knowledge among emails corpus. Some works have been done on emails related to topics classification or spam detection (Jindal, 2007), it’s rarely used in the context of knowledge management. The use of email is often confined to coordination and planning tasks (Wasiak et al, 2011), (Matta et al, 2010) or in case of legal issues. Enron case turned into a rich corpus for research about communication and social networks (Diesner, 2005). We focus in this paper on problem solving analysis e-mails. We use pragmatics analysis and speech act in order to identify corresponding sentence intentions by also linking speech acts to project context and organization. 2 DESIGN PROBLEM SOLVING Problem solving plays a central role in design projects. For instance, in software design, developer’s deal with needs’ analysis, specification documents, implementation, debugging and testing. This is especially true if the development follow an agile method, with several round trips from design to delivery. According to Hardin (Hardin, 2002), “Any problem has at least three components: § Givens: information and facts presenting context; § Goal: desired end state; § Operations: actions to be performed to reach end state; Software development evolved quickly in the last two decade from classic waterfall model to Extreme Programming and currently to cyclic and iterative methods like Agile (Beck et al, 2001). As a side effect round trips between product-owner (contractor) and product-manager (developer) are more frequent, leading to increased communication and collaborative work. Typically, problem Solving sequences happen on weekly (sometime daily) basis and imply all the actors of the project, not only the development team. Abstract: Computer mediated communication is ubiquitous in Software design projects. Email is used for project coordination, but also for design, implementation and test. Especially with currents agile development methods, it is very common to interact through computer meditated communication like email, instant messaging and other collaborative tools in order to express functional needs, notify of issues and take appropriate decisions. In this paper we propose a Knowledge Trace 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 practices 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. KTR: an approach that supports Knowledge extraction from design interactions Rauscher Francois, Matta Nada and Atifi Hassan Institute ICD/Tech-CICO, University of Technology of Troyes, 12 rue Marie Curie, BP. 2060, 10010 Troyes Cedex, France, { francois.rauscher, nada.matta, hassan.atifi}@utt.fr