Machine Learning and Lightweight Semantics to Improve Enterprise Search and Knowledge Management

Abstract

Enterprise search and knowledge management tools are amongst the most widely used tools within enterprises. Despite the breadth of tasks they are used for, and the large number of users they are used by, current tools supporting enterprise search are fairly generic and provide the same results to everyone. This chapter describes approaches using machine learning and lightweight semantics to make generic knowledge management tools context and task sensitive with the goal of increasing the productivity of knowledge workers. We focus on two key enterprise problems: (1) Enterprise search and (2) Collaborative Document Development. We describe our work on augmenting an enterprise search tool with context mining, process mining, and visualization technologies that use a combination of bag of words and lightweight semantic representations, making users more productive and efficient. We also describe our efforts on two enterprise tools to help improve the effectiveness and efficiency of knowledge workers collaboratively creating documents. All three of these were tested extensively at Accenture and two have been deployed and made available to over 150,000 Accenture employees. We present evaluation results using usage logs as well as questionnaires showing that these prototypes are effective at making consultants at Accenture more efficient as well as helping them find better information while being easier to use than existing enterprise tools.