I’ve been asked a few times whether the abstracts for the upcoming presentations at Search Solutions 2010 are available online anywhere. Normally, we’d just provide these as part of the proceedings for the event itself (in hard copy form), but I don’t see any harm in making these more widely available in advance. So what follows is a sneak peek for all 12 presentations plus the panel session. Further details of the event itself, incuding how to register, can be found on the IRSG website.
- Behshad Behzadi, Google: “Web Search Freshness” Freshness (timeliness) of web search results matters for a significant number of Web Search Queries. Web Search engines have to account for the temporal aspects of queries in order to show the most relevant results to their users. In this talk, I will describe the criticality of Freshness in Web Search. I will give an overview of the different levels of Freshness which have to be considered to improve the user experience. I will also talk about the challenges that we face in improving Freshness.
- Vishwa Vinay, Microsoft: “Click evidence – Signals and Tasks”Using Enterprise Search as the backdrop, this talk will attempt to cover the myriad virtues of the click event. Seen as an implicit feedback signal from users, there has been much work describing a wide range of applications where such a signal is useful. Examples include ranking features, merging results from federated sources, evaluation, and obtaining related queries. While plentiful and relatively easy to obtain, the challenge with making the best use of click data is often in distinguishing the noise from the signal. This necessitates models and assumptions about the behaviour of users, a better understanding of our users in turn provides the basis for improvements in search systems. I will review some such models, with a bias towards published work coming out of Microsoft Research.
- Vivian Lin Dufour, Yahoo: “How to help searchers become better searchers” “How do we help users survey several aspects of their topic? How do we help users discover queries that will retrieve trending topics? This talk introduces innovative techniques of a directive search engine that helps the autonomous user at query session and intersession levels to discover and survey as many results as possible in the shortest amount of time.”
- Nick Patience, 451 Group: “The trends shaping the future of enterprise search 2010-2013″ The presentation will focus on the influences that will shape the enterprise search market over the next three years. Information governance, search-based applications and other topics will be examined and analyzed and some predictions made for where things will be in 2013.
- Chirag Gandhi, mPhasis: “I Still haven’t found what I am looking for…” After spending millions of $$$, clients realize they still cannot find what they are looking for… The features do not work as expected and are not meeting requirements. The project team is up against the wall with a looming deadline and unhappy users who are unable to view the appropriate search results… a no-go situation. We’ll showcase a couple of cases where the search engine let us down; how we managed to survive and came out better.
- Dusan Rnic, Endeca: “Enterprise Search and its Evolution” Since the early 60’s, traditional search has provided users with a method for finding information, with specific tasks at hand. As the volume, variety, and complexity of information explodes exponentially, traditional methods are no longer able to keep up with the demands. In an age of information, how do you encourage information discovery, leading to more informed decisions. The methods for finding information must also take into consideration the unpredictable; the Black Swan. This is where traditional search fails. A discovery process, based in user experience and driven by increased information visibility is key. Fostering individual discretion for information discovery is key to encouraging strategic thinking, increasing revenue and profits. Join us for a journey through information discovery, looking back to the foundations of traditional search through to methods of information visibility and access, designed to provide users with the actionable intelligence.
- Greg Lindahl, Blekko: “Instant Indexing” Web-scale search engines have some interesting challenges standing in the way of providing features such as quick addition of newly-crawled material into the index. The desire to answer queries in 100 milliseconds or less leads to highly compressed indexes which are difficult to update. In addition, a high rate of updates to a small number of items in the index can cause hotspot problems. In this talk, I will describe an incremental index update technique using combinable transactional operations, which we call ‘combinators’. I will contrast this with the transactions/notifications method described in a recently-published Google Percolator paper.
- Charlie Hull, Flax: “What’s the story with open source? — Searching and monitoring news media with open-source technology” We’re drowning in news – surrounded by newspapers, magazines and blogs and bombarded by email alerts, e-zines and RSS feeds – to say nothing of the tide of social media updates. Using case studies and examples from leading UK news organisations, Charlie will discuss the challenges presented by high volume news content and how to implement accurate and highly scalable search systems for news, using open source technology.
- Roberto Cornacchia, Spinque: “Search by strategy” Spinque promotes search by strategy, an iterative 2-stage search process that separates search strategy definition (the how) from the actual searching and browsing of the collection (the what). Search by strategy allows information specialists to model, test and refine their search workflows at a high level of abstraction, with the help of a graphical tool. Any intermediate results can be inspected, keeping search transparent and the information specialist in control. Successful search strategies can be kept for future re-use, or deployed as an automatically generated search engine for end users, which runs on a relational database system. The magic that enables a fully automatic processing of the resulting data flow is found in our specific take on mixing information retrieval theory and databases.
- Till Kinstler, German Common Library Network: “Current trends in library search: From electronic card boxes to large scale, aggregated search engines” Traditional library search interfaces like Online Public Access Catalogues (OPACs) offer access to only a small subset of library collections (mainly books), provide limited search capabilities and usually show search results ordered by a single aspect of bibliographic data (title, author, date…). Metasearch engines, which were implemented in libraries to provide federated access to additional article and subject databases through a single search interface, are not liked by library users because of usability issues. Such issues are now being addressed by the implementation of large aggregated search indexes that aim at providing integrated access to all content and services available at a library. The data for these indexes is aggregated from various sources, like traditional library management systems, publishers or web sites. These indexes are usually based on search engine products like Solr/Lucene, Fast or Xapian, that implement standard information retrieval technology (vector space model, relevance ranking, flexible indexing capabilities etc.). The presentation will highlight some of the issues experienced in implementing such an aggregated search engine for the German Research Foundation’s (DFG) large-scale collection of nationally licensed digital material.
- Mihai Lupu, Information Retrieval Facility: “Scaling up innovation” We talk often in IR about the problem of scaling up to the amount of data which is being constantly generated, at an increasing pace, by people, for-profit or not-for-profit, governmental or non-governmental organisations. But can we speed up the innovation cycle? The IRF has been established 4 years ago with the specific purpose of being a platform between industry and academia, on top of which a sustainable innovation cycle could be developed and perfected. By providing standardised test collections and large computing facilities to researchers, it promotes a healthy competition,but also a deep understanding of search technologies and their specific benefits for each case study. Through its pool of experts, industry members can take advantage of this knowledge easily and directly support further research. Its success in attracting governmental and European funding further shows that the problems addressed are of significant interest not only commercially but also for the community. This talk will provide a summary of the IRF’s activities and invites the listeners to be part of the accelerating innovation cycle.
- Rob Stacey, TrueKnowledge: “Reconciling facts: how to check the consistency of facts created from web crawling” “Structured semantic information can be constructed from the unstructured web, natural language can be turned into facts in isolation but how are these reconciled with a larger knowledge base that may contain similar or contradictory information? By using True Knowledge’s Answer Engine to assess the information, facts constructed by crawling the internet can be added to the knowledge base without creating contradictions or semantic conflicts. True Knowledge uses its Answer Engine to validate a given fact and in doing so provides the system with information on the truth of the fact and whether the fact can already be inferred by the system. This process is called system assessment and uses True Knowledge’s full Knowledge Base of over 300 millions facts to provide reasoning on the validity of crawled information that can be used to determine why a piece of information is considered correct or incorrect by the system. This allows True Knowledge to add facts generated from web crawling into its growing knowledge base and enables questions to be answered using them.
What will search look like in 2015?
One of the interesting things about Yahoo’s recently launched Time Explorer is that it “lets users explore the future” and “allows users to view predictions that are yet to occur.” So that raises the question – what does it say about the way search will look in five years time? Unfortunately, the corpus it uses doesn’t provide detailed answers to questions like this just yet.
But we can pick up the baton. With Search Solutions entering its 5th successful year, it is a timely opportunity to ask the question ourselves: What will search look like in five years time? Which startups are most likely to change the commercial landscape? Which research ideas and prototypes will have become mainstream? If you want to find answers to questions such as these, then come and join us for the panel session at Search Solutions 2010.
- Nick Patience, 451 Group
- Stefan Rueger, Open University
- Jody Goodall, Trader Media
- Charlie Hull, Flax