The PwC Connection Machine: An Adaptive Expertise Provider

The Connection Machine helps PricewaterhouseCoopers LLP (PwC) partners and staff to solve problems by connecting people to people. It allows information seekers to enter their question in free text, finds knowledgeable colleagues, forwards the question to them, obtains the answer and sends it back to the seeker. In the course of this interaction, the application unobtrusively learns and updates user profiles and thereby increases its routing accuracy. The Connection Machine combines features of expertise locators, adaptive casebased recommender systems and question answering applications. This document describes the core technology that supports the workflow, the user modeling and the retrieval technology of the Connection Machine. 1 The Power of Connected People Information, knowledge and experience are key success factors and the most important competitive advantage for any business. However, most of this core corporate asset is in the heads of the employees and cannot be easily accessed, shared or distributed. Capturing and protecting it in documents (electronic or otherwise) is not only cumbersome, but the documents become rapidly outdated and the maintenance effort required to keep document collections up-to-date is formidable. Furthermore, in the complex business scenarios of today’s world, problem solving requires an increasingly large amount of specialized knowledge. It is nearly impossible for one individual to be an expert in every aspect of a company’s business and deliver comprehensive solutions. Problem solving requires co-operation and the sharing of ideas and information. The size of a corporation and the collective knowledge of its employees are only valuable if these employees can share their information and cooperate. We believe that the best way to provide the most up-todate and accurate information to those who seek it is by putting them directly in touch with the experts. The PricewaterhouseCoopers Connection Machine is an application that enables employees to solve business problems by helping them obtain answers to their questions from knowledgeable colleagues. Rather than trying to extract information from experts and pointing information seekers to stale document directories, the Connection Machine matches incoming questions to the expertise profiles of users, routes questions to the experts with highest similarity, collects their answers and relays the answer back to the seekers. The application extends the personal network of employees to the entire firm and makes otherwise difficult to reach experts accessible. 2 Existing Approaches to Locating and Contacting Experts 2.1 Directory Systems Most firms allow their employees to search for other colleagues by means of directories. Typically, these directories list the business unit, office phone numbers and addresses of employees, as well as some limited information about their background. Searches are usually performed by entering the (partial) name of the employee or by browsing to through the business unit structure of the firm. In terms of their functionality, these systems resemble phone books with a job categorization, similar to “yellow pages”. If we know which employee we are looking for, directory systems are very useful for finding their contact information. However, most of these applications do not help to determine which employee might be knowledgeable on a specific topic [c.f. 1] and able or willing to answer our question. They also do not help to relay the question to the right person, to obtain an answer in a given timeframe, or to create a network of employees. Additionally, the data that goes beyond office location, department, phone numbers etc. is typically not centrally maintained and requires manual updates by the employees themselves. As such, the information is mostly outdated and its reliability rather limited. Also, in personal interactions, if experts are not able to give an answer to a question, they typically refer the inquirer to another specialist from their personal network. A user looking for an expert in a directory system has only access to one level of experts and is at the mercy of the expert he/she contacts. People who have no representative profile in the directory system are beyond the reach of the seeker entirely. Since standard directory applications do not provide the functionality to find an expert and ask a question easily, employees typically revert to the rather inefficient practice of sending emails to broad audiences in the hope of finding someone who is able and willing to help them. 2.2 Expertise Locator Systems To answer the need for being able to access experts and ask questions, companies have developed so called Expertise Locator Systems (ELS). These systems try to find experts that are potentially able to answer a user’s question by matching the query to the expertise profiles of the employees [2, 3, 4, 5]. Some systems enhance the matching process by using the social connections between employees or collaborative filtering (e.g. [1, 6]). Depending on the application, they return a combination of potentially knowledgeable experts and related documents. It is up to the user looking for information to contact the experts and to get an answer to their question Employees are normally represented by an expertise profile which, depending on the application, contains a limited number of structured attributes coming from an enterprise directory, a list of documents published by the employee, and general background information in free text or in a list of terms/noun phrases. The experts can update their profiles manually by adding new documents, modifying their background information and, potentially, the structured data. Responses to queries can be published and added to the profile as new documents as well. Some systems generate profiles automatically by analyzing emails and authored documents and extracting a set of terms. Users have to go through the terms to specify which ones represent areas that they would feel comfortable answering questions in. Current expertise locator systems are designed to search for people. They match the user’s question with documents and expert’s profiles and display the list of matching experts to the users. The users, in turn, have to pick an expert from this list and contact them with the question. However, the goal of users who submit questions to an expertise locator system is not to find the name of colleagues but to find answers to their questions! The fact that a user has found the name of a potentially knowledgeable person does not mean that his/her question has been answered. An additional weakness in current expertise locator systems is the lack of division between interest and expertise. Existing expertise locators analyze documents that have been authored by users and their emails to generate a profile to represent each user’s expertise. If a user subscribes to an electronic newsletter out of interest in the specific topic, or writes a "Request for Proposals" (RFP) for vendors to respond to, he/she will be presumed an expert in that field. 3 Overview of the PwC Connection Machine The Connection Machine extends the concepts of directory systems and expertise locators beyond the pure search for people and helps PwC partners and staff to get answers to their questions and to solve problems together. It leverages the personal networks and intelligence of PwC employees, facilitates collaborative problem solving, and fosters a work environment in which people are truly connected. By answering questions rather than just locating people, the Connection Machine acts as a virtual, adaptive expertise provider. It combines features of expertise locators with adaptive case-based recommender systems and question answering applications. Figure 1 provides a general overview of the interaction between the information Seeker, potential Providers and the Connection Machine. Figure 1: Overview of the application workflow in the PwC Connection Machine An interaction with the Connection Machine starts with an Information Seeker entering a question in free text format, as if he/she were asking a colleague a question via email. The Seeker is also able to specify the urgency of the question, the name of a client the question relates to as well as additional, optional, structured information (e.g. knowledge domain, line of service, industry) to be used to locate appropriate potential Providers (Figure 2). The Connection Machine processes the query, finds a set of suitable potential Providers and contacts them. The system only contacts potential Providers whose expertise levels for the given question are higher than the Seeker’s and whose maximum number of questions per week has not been reached. Figure 2: Web interface of the PwC Connection Machine The number of potential Providers who will be contacted regarding a question is configurable. If the first set of potential Providers is not able to respond within the allocated time (a fraction of the time the Seeker needs the answer by), the system sends the question to a second batch of potential Providers. If no potential Providers can be identified or the Providers do not react, the question is sent to the Knowledge Administrator of the Domain for further processing. Once the system identifies potential Providers, they are notified via email (Figure 3) and a visual indicator in the “Summary” page of the web interface, informing them that their expertise is needed. In addition to the question, the potential Providers are informed of the Seeker’s contact information (e.g. name, line of service) and of the timeframe in which the question needs to be answered. After receiving a question, the potential Providers may choose to respond either via web interface or via email. Potential Providers may offer an answer to the question; request additional information from the Seeker; refer the question to other potential Providers; or decline to answer. Once one of the potential Providers offers an answer or requests additional information, he/she becomes the “Provider” for the interaction. From this point on, the Connection Machine facilitates communicates between the information Seeker and the Provider and removes other potential Providers from the problem solving conversation by sending them email and removing the indicator in their “Summary” web page. Figure 3: Sample email from the PwC Connection Machine If a Provider chooses to answer the question, the Seeker is notified of the answer via email and a visual indicator in the web application (Figure 4). Upon receiving an answer, a Seeker can choose to accept it and close the request, ask a clarification question about the answer, or reject the answer and request a second opinion unless they have already done so. The Provider can also ask for additional information that may be needed to answer the question. Once the Seeker provides additional information the Provider will be presented with the same options as when initially contacted by the Connection Machine (i.e. provide an answer, request question clarification, refer the question and decline to answer). If a (potential) Provider decides that someone else from his/her personal network is better suited to answer the question, he/she may choose to refer the question. In this way the Connection Machine can learn about potential Providers who may have been missing from its initial set of profiles. The Seeker will not be made aware that the question was referred to another potential Provider as long as the initial Provider had not contacted the Seeker prior to referring the question (i.e. the provider did not request question clarification prior to referring the question). The Provider can also indicate that he/she is not able to provide an answer to the question and specify the reason for declining to answer (e.g. “Too busy”, “Don’t know the answer”, “Independence conflict”). If the question was declined by all contacted Providers, it will be sent to the Knowledge Administrator of the domain for further processing. The use of a Knowledge Administrator as a “backup” for answering or referring questions ensures that all questions entered in the Connection Machine are answered in a timely manner. Figure 4: Open question summary page 4 Retrieval of Potential Providers in the Connection Machine To execute the workflow described above the application needs to be able to determine who is potentially capable of answering the question by matching a user’s query against information it has about other users. To achieve consistently high accuracy over a long period of time, the information of the users has to be updated with appropriate sections of the interaction on an ongoing basis. (Figure 5) The technology we used to implement these functions in the Connection Machine is similar to User Adaptive, Case-Based Recommender Systems [7, 8, 9]. However, most recommender systems are geared towards selecting the best match out of a set of (mostly static) items and presenting it to the user. In the case of the Connection Machine, the items in the case-base are continuously evolving user models where each model contains multiple profiles. Rather than being the final goal, the retrieval process is an intermediate step and users, whose expertise profile matched the query, are utilized in the workflow to route questions. The resulting interaction between the Seeker and Provider is the desired outcome for the application. The user modeling in the Connection Machine is not geared towards influencing the similarity metrics, the user interaction or the user interface of the application. Neither can it influence the solutions a Provider may offer to a Seeker. The case-base of the Connection Machine is a collection of user models which are constantly maintained and updated by unobtrusively observing the user’s interaction with the system [cf. 10] and, which are then used to select the users that will participate in the workflow as potential Providers.