This Friday, we'll have two talks followed by drinks. The academic talk will be given by Christophe Van Gysel. He will talk about expertise retrieval (see the abstract below). Andreas Kohn from Collaborne (http://www.collaborne.com/) will give the industry talk.
Christophe Van Gysel(ILPS-UvA)--Unsupervised, Efficient and Semantic Expertise Retrieval
We introduce an unsupervised discriminative model for the task of retrieving experts in online document collections. We exclusively employ textual evidence and avoid explicit feature engineering by learning distributed word representations in an unsupervised way. We compare our model to state-of-the-art unsupervised statistical vector space and probabilistic generative approaches. Our proposed log-linear model achieves the retrieval performance levels of state-of-the-art document-centric methods with the low inference cost of so-called profile-centric approaches. It yields a statistically significant improved ranking over vector space and generative models in most cases, matching the performance of supervised methods on various benchmarks. That is, by using solely text we can do as well as methods that work with external evidence and/or relevance feedback. A contrastive analysis of rankings produced by discriminative and generative approaches shows that they have complementary strengths due to the ability of the unsupervised discriminative model to perform semantic matching.
Christophe is a PhD candidate at ILPS, advised by Maarten de Rijke and Evangelos Kanoulas. His research focuses on semantic matching in information retrieval. He is designing models that learn semantics from document collections and applies these to real-world retrieval settings, such as expertise finding, product search and session search.
Andreas Kohn(Collaborne)--Search and Beyond: Challenges for Search for Innovation Management
Collaborne provides a SaaS platform that makes innovation smart, simple, and sticky. Search in this environment is not limited to basic keyword search to let users find back content entered before. Instead, search is a fundamental building block to guide users through the innovation process, provide the right information to each user, or simply render dashboards. Search as a building block raises many interesting challenges. For example: Users constantly move between the physical world – like user interviews or brain storming sessions with Post-It notes – and the virtual world on their desktop and mobile devices. Users have also very diverse information needs: outsiders only enter "weak signals"; innovators create rich outcomes as they proceed from insights, via ideas, to prototypes; line managers monitor the project progress; and senior managers set the organizations' challenges and goals for continuous innovation. The talk looks into Collaborne's current use of search technology to tackle these challenges – and will outline potential future improvements.
Andreas Kohn has been working with search in enterprise software for the last ten years, starting as a developer in (SDL) Fredhopper, where he was in charge of improving the existing proprietary search solution to handle the changing customer expectations in e-commerce software. At Collaborne, he is working on the complete search stack – from operations in the AWS cloud, over implementing the Elasticsearch-based backend, to user experience in the frontend. He enjoys knowing that his work helps organizations like the City of Amsterdam, Vodafone, the National Police, BNP Paribas to innovate faster and better.