The University of Amsterdam at Trec 2010: Session, Entity, and Relevance Feedback

We describe the participation of the University of Amsterdam’s ILPS group in the session, entity, and relevance feedback track at TREC 2010. In the Session Track we explore the use of blind relevance feedback to bias a follow-up query towards or against the topics covered in documents returned to the user in response to the original query. In the Entity Track REF task we experiment with a window size parameter to limit the amount of context considered by the entity co-occurrence models and explore the use of Freebase for type filtering, entity normalization and homepage finding. In the ELC task we use an approach that uses the number of links shared between candidate and example entities to rank candidates. In the Relevance Feedback Track we experiment with a novel model that uses Wikipedia to expand the query language model.

  • [PDF] M. Bron, J. He, K. Hofmann, E. Meij, M. de Rijke, E. Tsagkias, and W. Weerkamp, “The University of Amsterdam at TREC 2010: session, entity, and relevance feedback,” in The nineteenth text retrieval conference, 2011.
    Abstract = {We describe the participation of the University of Amsterdam's Intelligent Systems Lab in the web track at TREC 2009. We participated in the adhoc and diversity task. We find that spam is an important issue in the ad hoc task and that Wikipedia-based heuristic optimization approaches help to boost the retrieval performance, which is assumed to potentially reduce spam in the top ranked results. As for the diversity task, we explored different methods. Clustering and a topic model-based approach have a similar performance and both are relatively better than a query log based approach.},
    Author = {M. Bron and He, J. and Hofmann, K. and Meij, E. and de Rijke, M. and Tsagkias, E. and Weerkamp, W.},
    Booktitle = {The Nineteenth Text REtrieval Conference},
    Date-Added = {2011-10-20 11:18:35 +0200},
    Date-Modified = {2012-10-30 09:25:06 +0000},
    Series = {TREC 2010},
    Title = {{The University of Amsterdam at TREC 2010}: Session, Entity, and Relevance Feedback},
    Year = {2011}}
thesis cover image of a smart computer

Combining Concepts and Language Models for Information Access

Since the middle of last century, information retrieval has gained an increasing interest. Since its inception, much research has been devoted to finding optimal ways of representing both documents and queries, as well as improving ways of matching one with the other. In cases where document annotations or explicit semantics are available, matching algorithms can be informed using the concept languages in which such semantics are usually defined. These algorithms are able to match queries and documents based on textual and semantic evidence.

Recent advances have enabled the use of rich query representations in the form of query language models. This, in turn, allows us to account for the language associated with concepts within the retrieval model in a principled and transparent manner. Developments in the semantic web community, such as the Linked Open Data cloud, have enabled the association of texts with concepts on a large scale. Taken together, these developments facilitate a move beyond manually assigned concepts in domain-specific contexts into the general domain.

This thesis investigates how one can improve information access by employing the actual use of concepts as measured by the language that people use when they discuss them. The main contribution is a set of models and methods that enable users to retrieve and access information on a conceptual level. Through extensive evaluations, a systematic exploration and thorough analysis of the experimental results of the proposed models is performed. Our empirical results show that a combination of top-down conceptual information and bottom-up statistical information obtains optimal performance on a variety of tasks and test collections.

See for more information.

  • [PDF] E. Meij, “Combining concepts and language models for information access,” PhD Thesis, 2010.
    Author = {Meij, Edgar},
    Date-Added = {2011-10-20 10:18:00 +0200},
    Date-Modified = {2011-10-22 12:23:33 +0200},
    School = {University of Amsterdam},
    Title = {Combining Concepts and Language Models for Information Access},
    Year = {2010}}



A query model based on normalized log-likelihood

A query is usually a brief, sometimes imprecise expression of an underlying information need . Examining how queries can be transformed to equivalent, potentially better queries is a theme of recurring interest to the information retrieval community. Such transformations include expansion of short queries to long queries, paraphrasing queries using an alternative vocabulary, mapping unstructured queries to structured ones, identifying key concepts in verbose queries, etc.

To inform the transformation process, multiple types of information sources have been considered. A recent one is search engine logs for query substitutions . Another recent example is where users complement their traditional keyword query with additional information, such as example documents, tags, images, categories, or their search history . The ultimate source of information for transforming a query, however, is the user, through relevance feedback : given a query and a set of judged documents for that query, how does a system take advantage of the judgments in order to transform the original query and retrieve more documents that will be useful to the user? As demonstrated by the recent launch of a dedicated relevance feedback track at TREC, we still lack the definitive answer to this question.

Let’s consider an example to see what aspects play a role in transforming a query based on judgments for a set of initially retrieved documents. Suppose we have a set of documents which are judged to be relevant to a query. These documents may vary in length and, furthermore, they need not be completely on topic because they may discuss more topics than the ones that are relevant to the query. While the users’ judgments are at the document level, not all of the documents’ sections can be assumed to be equally relevant. Most relevance feedback models that are currently available do not model or capture this phenomenon; instead, they attempt to transform the original query based on the full content of the documents. Clearly this is not ideal and we would like to account for the possibly multi-faceted character of documents. We hypothesize that a relevance feedback model that attempts to capture the topical structure of individual judged documents (“For each judged document, what is important about it?”) as well as of the set of all judged documents (“Which topics are shared by the entire set of judged documents?”) will outperform relevance feedback models that capture only one of these types of information.

We are working in a language modeling (LM) setting and our aim in this paper is to present an LM-based relevance feedback model that uses both types of information—about the topical relevance of a document and about the general topic of the set of relevant documents— to transform the original query. The proposed model uses the whole set of relevance assessments to determine how much each document that has been judged relevant should contribute to the query transformation. We use the TREC relevance feedback track test collection to evaluate our model and compare it to other, established relevance feedback methods. We show that it is able to achieve superior performance over all evaluated models. We answer the following two research questions in this paper. (i) Can we develop a relevance feedback model that uses evidence from both the individual relevant documents and the set of relevant documents as a whole? (ii) Can our new model achieve state-of-the-art results and how do these results compare to related models? When evaluated, we show that our model is able to significantly improve over state-of-art feedback methods.

  • [PDF] E. Meij, W. Weerkamp, and M. de Rijke, “A query model based on normalized log-likelihood,” in Proceedings of the 18th acm conference on information and knowledge management, 2009.
    Author = {Meij, Edgar and Weerkamp, Wouter and de Rijke, Maarten},
    Booktitle = {Proceedings of the 18th ACM conference on Information and knowledge management},
    Date-Added = {2011-10-12 18:31:55 +0200},
    Date-Modified = {2012-10-30 08:42:51 +0000},
    Series = {CIKM 2009},
    Title = {A query model based on normalized log-likelihood},
    Year = {2009},
    Bdsk-Url-1 = {}}
Stack of books

Concept models for domain-specific search

We describe our participation in the 2008 CLEF Domain-specific track. We evaluate blind relevance feedback models and concept models on the CLEF domain-specific test collection. Applying relevance modeling techniques is found to have a positive effect on the 2008 topic set, in terms of mean average precision and precision@10. Applying concept models for blind relevance feedback, results in even bigger improvements over a query-likelihood baseline, in terms of mean average precision and early precision.

  • [PDF] E. Meij and M. de Rijke, “Concept models for domain-specific search,” in Evaluating systems for multilingual and multimodal information access, 9th workshop of the cross-language evaluation forum, clef 2008, aarhus, denmark, september 17-19, 2008, revised selected papers, 2009.
    Author = {Meij, Edgar and de Rijke, Maarten},
    Booktitle = {Evaluating Systems for Multilingual and Multimodal Information Access, 9th Workshop of the Cross-Language Evaluation Forum, CLEF 2008, Aarhus, Denmark, September 17-19, 2008, Revised Selected Papers},
    Date-Added = {2011-10-12 18:31:55 +0200},
    Date-Modified = {2012-10-30 08:44:35 +0000},
    Title = {Concept models for domain-specific search},
    Year = {2009}}
CLEF domain-specific sample graphic

The University of Amsterdam at the CLEF 2008 Domain Specific Track – Parsimonious Relevance and Concept Models

We describe our participation in the CLEF 2008 Domain Specific track. The research questions we address are threefold: (i) what are the effects of estimating and applying relevance models to the domain specific collection used at CLEF 2008, (ii) what are the results of parsimonizing these relevance models, and (iii) what are the results of applying concept models for blind relevance feedback? Parsimonization is a technique by which the term probabilities in a language model may be re-estimated based on a comparison with a reference model, making the resulting model more sparse and to the point. Concept models are term distributions over vocabulary terms, based on the language associated with concepts in a thesaurus or ontology and are estimated using the documents which are annotated with concepts. Concept models may be used for blind relevance feedback, by first translating a query to concepts and then back to query terms. We find that applying relevance models helps significantly for the current test collection, in terms of both mean average precision and early precision. Moreover, parsimonizing the relevance models helps mean average precision on title-only queries and early precision on title+narrative queries. Our concept models are able to significantly outperform a baseline query-likelihood run, both in terms of mean average precision and early precision on both title-only and title+narrative queries.

  • [PDF] E. Meij and M. de Rijke, “The University of Amsterdam at the CLEF 2008 Domain Specific Track – parsimonious relevance and concept models,” in Working notes for the clef 2008 workshop, 2008.
    Author = {Edgar Meij and Maarten de Rijke},
    Booktitle = {Working Notes for the CLEF 2008 Workshop},
    Date-Added = {2011-10-12 18:31:55 +0200},
    Date-Modified = {2012-10-30 09:28:58 +0000},
    Title = {The {U}niversity of {A}msterdam at the {CLEF} 2008 {Domain Specific Track} - Parsimonious Relevance and Concept Models},
    Year = {2008}}


Parsimonious Relevance Models

Relevance feedback is often applied to better capture a user’s information need. Automatically reformulating queries (or blind relevance feedback) entails looking at the terms in some set of (pseudo-)relevant documents and selecting the most informative ones with respect to the set or the collection. These terms may then be reweighed based on information pertinent to the query or the documents and—in a language modeling setting—be used to estimate a query model, P(t|θQ), i.e., a distribution over terms t for a given query Q.

Not all of the terms obtained using blind relevance feedback are equally informative given the query, even after reweighing. Some may be common terms, whilst others may describe the general domain of interest. We hypothesize that refining the results of blind relevance feedback, using a technique called parsimonious language modeling, will improve retrieval effectiveness. Hiemstra et al. already provide a mechanism for incorporating (parsimonious) blind relevance feedback, by viewing it as a three component mixture model of document, set of feedback documents, and collection. Our approach is more straightforward, since it considers each feedback document separately and, hence, does not require the additional mixture model parameter. To create parsimonious language models we use an EM algorithm to update the maximum-likelihood (ML) estimates. Zhai and Lafferty already proposed an approach which uses a similar EM algorithm; it differs, however, in the way the set of feedback documents is handled. Whereas we parsimonize each individual document, they apply their EM algorithm to the entire set of feedback documents.

To verify our hypothesis, we use a specific instance of blind relevance feedback, namely relevance modeling (RM). We choose this particular method because it has been shown to achieve state-of-the-art retrieval performance. Relevance modeling assumes that the query and the set of documents are samples from an underlying term distribution—the relevance model. Lavrenko and Croft formulate two ways of approaching the estimation of the parameters of this model. We build upon their work and compare the results of our proposed parsimonious relevance models with RMs as well as with a query-likelihood baseline. To measure the effects in different contexts, we employ five test collections taken from the TREC-7, TREC Robust, Genomics, Blog, and Enterprise tracks and show that our proposed model improves performance in terms of mean average precision on all the topic sets over both a query-likelihood baseline as well as a run based on relevance models. Moreover, although blind relevance feedback is mainly a recall enhancing technique, we observe that parsimonious relevance models (unlike their non-parsimonized counterparts) can also improve early precision and reciprocal rank of the first relevant result. Thus, our parsimonious relevance models (i) improve retrieval effectiveness in terms of MAP on all collections, (ii) significantly outperform their non-parsimonious counterparts on most measures, and (iii) have a precision enhancing effect, unlike other blind relevance feedback methods.

  • [PDF] E. Meij, W. Weerkamp, K. Balog, and M. de Rijke, “Parsimonious relevance models,” in Proceedings of the 31st annual international acm sigir conference on research and development in information retrieval, 2008.
    Author = {Meij, Edgar and Weerkamp, Wouter and Balog, Krisztian and de Rijke, Maarten},
    Booktitle = {Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval},
    Date-Added = {2011-10-12 18:31:55 +0200},
    Date-Modified = {2012-10-30 08:47:44 +0000},
    Series = {SIGIR 2008},
    Title = {Parsimonious relevance models},
    Year = {2008},
    Bdsk-Url-1 = {}}

Thesaurus-Based Feedback to Support Mixed Search and Browsing Environments

We propose and evaluate a query expansion mechanism that supports searching and browsing in collections of annotated documents. Based on generative language models, our feedback mechanism uses document-level annotations to bias the generation of expansion terms and to generate browsing suggestions in the form of concepts selected from a controlled vocabulary (as typically used in digital library settings). We provide a detailed formalization of our feedback mechanism and evaluate its effectiveness using the TREC 2006 Genomics track test set. As to the retrieval effectiveness, we find a 20% improvement in mean average precision over a query-likelihood baseline, whilst increasing precision at 10. When we base the parameter estimation and feedback generation of our algorithm on a large corpus, we also find an improvement over state-of-the-art relevance models. The browsing suggestions are assessed along two dimensions: relevancy and specificity. We present an account of per-topic results, which helps understand for what type of queries our feedback mechanism is particularly helpful.

  • [PDF] E. Meij and M. de Rijke, “Thesaurus-based feedback to support mixed search and browsing environments,” in Research and advanced technology for digital libraries, 11th european conference, ecdl 2007, 2007.
    Author = {Edgar Meij and Maarten de Rijke},
    Booktitle = {Research and Advanced Technology for Digital Libraries, 11th European Conference, ECDL 2007},
    Date-Added = {2011-10-12 18:31:55 +0200},
    Date-Modified = {2012-10-28 23:04:22 +0000},
    Title = {Thesaurus-Based Feedback to Support Mixed Search and Browsing Environments},
    Year = {2007}}