Postdoc position in Semantic Entity Search

About half a year ago I advertised a PhD position in Semantic Entity Search. There were no eligible candidates, so it has been converted to a 2-year Postdoc position.

There is good flexibility topic-wise—as long as it’s about entities and semantics :)
Please feel free to contact me with any questions or for further information.

Details and application instructions can be found here.

You will notice that there is a number of projects advertised. It’s a department-funded position, so “may the best applicant win” is the name of the game. Meaning, the strongest candidate will be offered the position, irrespective of the project chosen.

Starting date: from Sept 1, 2014.
Application deadline: June 22, 2014.

Living Labs for Information Retrieval Evaluation

Evaluation is a central aspect of information retrieval (IR) research. In the past few years, a new evaluation methodology known as living labs has been proposed as a way for researchers to be able to perform in-situ evaluation. This is not new, you might say; major web search engines have been doing it for serveral years already. While this is very true, it also means that this type of experimentation, with real users performing tasks using real-world applications, is only available to those selected few who are involved with the research labs of these organizations. There has been a lot of complaining about the “data divide” between industry and academia; living labs might be a way to bridge that.

The Living Labs for Information Retrieval Evaluation (LL’13) workshop at CIKM last year was a first attempt to bring people, both from academia and industry, together to discuss challenges and to formulate practical next steps. The workshop was successful in identifying and documenting possible further directions. See the preprint of the workshop summary.

The second edition of the iving Labs for IR workshop (LL’14), will run at CIKM this year. Our main goals are to continue our community building efforts around living labs for IR and to pursue the directions set out at LL’13. Having a community benchmarking platform with shared tasks would be a key catalyst in enabling people to make progress in this area. This is exactly what we are trying to set up for LL’14, in the form of a challenge (with the ultimate goal of turning it into a TREC, NTCIR or CLEF track in the future).

The challenge focuses on two specific use-cases: product search and local domain search. The basic idea is that participants receive a set of 100 frequent queries along with candidate results for these queries, and some general collection statistics. They are then expected to produce rankings for each query and to upload these rankings through an API. These rankings are evaluated online, on real users, and the results of these evaluations are made available to the participants, again, through an API.

In preparation for this challenge, we are organising a challenge workshop in Amsterdam on the 6th of June. The programme includes invited talks and a “hackathon.” We have a limited number of travel grants available (for those coming from outside The Netherlands and coming from academia) to cover travel and accommodation expenses. These are available on a “first come first served” basis (at most one per institute). If you would like to make use of this opportunity, please let us know as soon as possible.

More details may be found on our brand-new website: living-labs.net.

Temporal Expertise Profiling

Expertise is not a static concept. Personal interest as well as the landscape of respective fields change over time; knowledge becomes outdated, new topics emerge, and so on.
In recent work, Jan Rybak, Kjetil Nørvåg, and I have been working on capturing, modeling, and characterizing the changes in a person’s expertise over time.

The basic idea that we presented in an ECIR’14 short paper is the following. The expertise of an individual is modelled as a series of profile snapshots. Each profile spanshot is a weighted tree; the hierarchy represents the taxonomy of expertise areas and the weights reflect the person’s knowledge on the corresponding topic. By displaying a series of profile snapshots on a timeline, we can have a complete overview of the development of expertise over time. In addition, we identify and characterize important changes that occur in these profiles. See our colorful poster for an illustration.

In an upcoming SIGIR’14 demo we introduce a web-based system, ExperTime, where we implemented these ideas. While our approach is generic, the system is particular to the computer science domain. Specifically, we use publications from DBLP, classified according to the ACM 1998 Computing Classification System. Jan also created a short video that explains the underlying ideas and introduces the main features of the system:

The next step on our research agenda is the evaluation of temporal expertise profiles. This is a challenging problem for two reasons: (1) the notions of focus and topic changes are subjective and are likely to vary from person to person, and (2) the complexity of the task is beyond the point where TREC-like benchmark evaluations are feasible. The feedback we plan to obtain with the ExperTime system, both implicit and explicit, will provide invaluable information to guide the development of appropriate evaluation methodology.

If you are interested in your temporal expertise profile, you are kindly invited to sign up and claim it. Or, it might already be ready and waiting for you: http://bit.ly/expertime.

PhD position in Semantic Entity Search

The University of Stavanger invites applications for a three-year doctorate scholarship in Information Technology, at the Faculty of Science and Technology, in the Department of Electrical Engineering and Computer Science, beginning September 1, 2014.

Project: Semantic Entity Search
Semantic search refers to the idea that the search engine understands the concepts, meaning and intent behind the query that the user enters into the search box, and provides rich and focused responses (as opposed to merely a list of documents). Entities, such as people, organizations or products, play a central role in this context; they reflect the way humans think and organize information. We can observe that major search engines (like Google or Apple’s SIRI) are becoming “smarter” day by day in recognizing specific types of objects (for example, locations, events or celebrities); yet, true semantic search has still a long way to go.
This project aims to develop a theoretically sound and computationally efficient framework for entity-oriented information access: the search and discovery of entities and relationships between entities. A key element to a successful approach is the combination of massive volumes of structured and unstructured information from the Document Web and the Data Web, respectively. Successful candidates will be expected to conduct research, design, develop, and deploy state-of-art, scalable information retrieval, information extraction and machine learning techniques for innovative entity-oriented search applications. The project will include both theoretical and empirical explorations, where lab-based results will be evaluated in ‘live’ environments with real users.

Qualifications: M.Sc. in Computer Science, Computational Linguistics, Mathematics or related fields by the appointment date. Good written and spoken command of English. Research experience or a track record of project based work, demonstrable interest in the domain, solid programming skills (particularly Java), and experience in manipulating and analyzing large data sets (esp. using Hadoop) are a clear plus.

The research fellow is salaried according to the State Salary Code, l.pl 17.515, code 1017, LR 20, ltr 50, of NOK 421 100,- per annum.

Details and application instructions can be found here.
Application deadline: January 11, 2014.

A Test Collection for Entity Search in DBpedia

With this SIGIR ’13 short paper, we try to address some of the action points that were identified as as important priorities for entity-oriented and semantic search at the JIWES workshop held at SIGIR ’12 (see the detailed workshop report). Namely: (A1) Getting more representative information needs and favoring long queries over short ones. (A2) Limiting search to a smaller, fixed set of entity types (as opposed to arbitrary types of entities). (A3) Using test collections that integrate both structured and unstructured information about entities.

An IR test collection has three main ingredients: a data collection, a set of queries, and corresponding relevance judgments. We propose to use DBpedia as the data collection; DBpedia is a community effort to extract structured information from Wikipedia. It is one of the most comprehensive knowledge bases on the web, describing 3.64M entities (in version 3.7). We took entity-oriented queries from a number of benchmarking evaluation campaigns, synthesized them into a single query set, and mapped known relevant answers to DBpedia. This mapping involved a series of not-too-exciting yet necessary data cleansing steps, such as normalizing URIs, replacing redirects, removing duplicates, and filtering out non-entity results. In the end, we have 485 queries with an average of 27 relevant entities per query.

Now, let’s see how this relates to the action points outlined above. (A1) We consider a broad range of information needs, ranging from short keyword queries to natural language questions. The average query length, computed over the whole query set, is 5.3 terms—more than double the length of typical web search queries (which is around 2.4 terms). (A2) DBpedia has a consistent ontology comprising of 320 classes, organized into a 6 levels deep hierarchy; this allows for the incorporation of type information at different granularities. (A3) As DBpedia is extracted from Wikipedia, there is more textual content available for those who wish to combine structured and unstructured information about entities.

The paper also includes a set of baseline results using variants of two popular retrieval models: language models and BM25. We found that the various query sub-sets (originating from different benchmarking campaigns) exhibit different levels of difficulty—this was expected. What was rather surprising, however, is that none of the more advanced multi-field variants could really improve over the simplest possible single-field approach. We observed that a large number of topics were affected, but the number of topics helped/hurt was about the same. The breakdowns by various query-subsets also suggest that there is no one-size-fits-all way to effectively address all types of information needs represented in this collection. This phenomenon could give rise to novel approaches in the future; for example, one could first identify the type of the query and then choose the retrieval model accordingly.

The resources developed as part of this study are made available here. You are also welcome to check out the poster I presented at SIGIR ’13.
If you have (or planning to have) a paper that uses this collection, I would be happy to hear about it!