Site Search Using Profile-Based Document Summarisation

Abstract

Text summarisation is the process of distilling the most important information from a source to produce an abridged version for a particular user or task. This demo presents the use of pro le-based summarisation to provide contextualisation and interactive support for site search and enterprise search. We employ log analysis to acquire continuously updated pro les to provide pro le-based summarisations of search results. These pro les could be capturing an individual's interests or those of a group of users. Here we look at acquiring pro les for groups of users.

MOTIVATION

Summarisation is a broad area of research [ 8 ]. The sort of information contained in a summary di ers according to the mechanism used in the summarisation process: It may highlight the basic idea (generic summarisation), or it may highlight the speci c user's individual area of interest (personalised summarisation). One of the techniques used to achieve personalisation is user pro ling. User pro les may include the preferences or interests of a single user or a group of users and may also include demographic information [ 4 ]. Normally, a user pro le contains topics of interest to that single user. We are interested in capturing pro les not of single but groups of users.

We utilise query and click logs to acquire a pro le reecting the population's search patterns and this pro le is being automatically updated in a continuous learning cycle. We are then applying the acquired pro les in the summarisation process to support users searching a document collection. The potential of personalised summarisation over generic summaries has already been demonstrated, e.g. [ 3 ], but summarisation of Web documents is typically based on the query rather than a full pro le, e.g. [ 11, 9 ]. Our speci c interest lies in enterprise search which is di erent from Web search and has attracted less attention [ 5 ]. The bene t of this context is that we can expect a more homogeneous population of searchers who are likely to share interests and Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

DIR 2013, April 26, 2013, Delft, The Netherlands. . information needs. Our hypothesis is that pro le-based summarisation can help a user in this process and guide the user to the right documents more easily (e.g. by presenting the summaries instead of or alongside snippets). 2.

METHODS AND EXAMPLES

The demo presents an integrated Solr-based search system applying a number of di erent methods for building summaries for search results. The rst two algorithms were designed for traditional (generic) summarisation, and they represent widely used baselines, e.g. [ 12 ]. The other three are all variations of an approach that has been proposed in the literature for building an adaptive community prole/domain model, a "biologically inspired model based on ant colony optimisation applied to query logs as an adaptive learning process" [ 1 ]. The approach is simple to implement, the idea here is that query logs are segmented into sessions and then turned into a graph structure. Figure 1 gives an example of part of the pro le as it has been derived from our query logs. We used the log les collected on the existing search engine over a period of three years1 to bootstrap this ant colony optimisation (ACO) model, i.e. our pro le. The example illustrates the domain-speci c nature of the derived pro les, e.g. the University library is named after Albert Sloman.

A pro le-based (extractive) summary of a document is then generated by turning the pro le into a at list of terms (we use three di erent methods to do this as explained further down) and selecting those sentences from the document 1More than 1.5 million queries, described in more detail elsewhere [ 6 ] that are most similar to the pro le using cosine similarity. Figure 2 shows an architectural diagram for our pro le-based summarisation system. Following DUC 2002 convention we select 100-word abstracts [ 7 ]. This gives us the following ve methods: 1. Random: Selects sentences from the document randomly [ 12 ]. 2. Centroid: A centroid-based approach to summarisation [ 10 ]. This algorithm takes into account rstsentence overlap and positional value, which are then used to generate a coherent summary. 3. ACO: A query graph built by processing the log data according to [ 1 ]. The entire model is turned into a at list of terms for summarisation. 4. ACO trimmed: Starting with ACO we trim all those edges whose weights fall below the overall average weight of an edge. The remaining model is turned into a at list of terms for summarisation. 5. ACO query re nements: The list of terms used for summarisation are all those that are directly linked to the query node in the ACO model.

Note that ACO and ACO trimmed are query-independent as they are using the entire model to generate the summary, whereas ACO query re nements is query-speci c (for the frequently submitted query \library" the selected terms are library, albert sloman library, library homepage, library opening times and catalogue, see Figure 1).

To illustrate the di erent summaries obtained using three of the summarisation methods we apply the methods to the University of Essex Library homepage2 and get the following summaries: Random: We provide you with rst class library facilities to complement and assist your studies. The Library allows you to access more than one million books, periodicals and micro lms. Our collections covering Latin America, Russia and Eastern Europe are of national signi cance. Viewing facilities for DVDs and videos are also available.

Centroid: We provide you with rst class library facilities to complement and assist your studies. You can nd out more by accessing our University library services. Our Albert Sloman Library is just a few minutes' walk from teaching buildings and student accommodation at our Essex Campus.

ACO query re nements: In addition, 110 networked PCs and terminals provide access to over 47,000 online journals, databases, e-books and library catalogues. Students at our Essex Campus can visit the Albert Sloman Library or borrow books from its collection via a daily dispatch service. The Albert Sloman Library has long opening hours, a total of 84 hours over seven days a week during term and 42.5 to 84 hours in vacations.

Obviously, the actual usefulness of such summaries can only be assessed in a realistic search setting. In a pilot study we found that the ACO-based summaries have the potential of outperforming the di erent baselines [ 2 ]. A task-based evaluation using TREC Interactive Track guidelines is currently being conducted. As the immediate next step, we are interested in investigating how the pro le can be integrated into multi-document summarisation.

References

[1] M-D. Albakour , U. Kruschwitz , N. Nanas , D. Song , M. Fasli , and A. De Roeck . Exploring ant colony optimisation for adaptive interactive search . In Proceedings of ICTIR , pages 213 { 224 . Springer, 2011 .

[2] A. Alhindi , U. Kruschwitz , and C. Fox . A pilot study on using pro le-based summarisation for interactive search assistance . In Proceedings of ECIR , pages 672 { 675 , 2013 .

[3] A. D az and P. Gervas. User-model based personalized summarization . Information Processing & Management , 43 ( 6 ): 1715 { 1734 , 2007 .

[4] S. Gauch , M. Speretta , A. Chandramouli , and A. Micarelli. User pro les for personalized information access . The Adaptive Web , pages 54 { 89 , 2007 .

[5] D. Hawking . Enterprise Search. In R. Baeza-Yates and B. Ribeiro -Neto, editors, Modern Information Retrieval , pages 641 { 683 . Addison-Wesley , 2nd edition , 2011 .

[6] U. Kruschwitz , D. Lungley , M-D. Albakour , and D. Song . Deriving Query Suggestions for Site Search . JASIST, 2013 . Forthcoming.

[7] C.Y. Lin and E. Hovy . Automatic evaluation of summaries using n-gram co-occurrence statistics . In Proceedings of HLT-NAACL , pages 71 { 78 . ACL , 2003 .

[8] A. Nenkova and K. McKeown . Automatic summarization . Now Publishers, 2011 .

[9] S. Park . Personalized summarization agent using non-negative matrix factorization . PRICAI 2008: Trends in Arti cial Intelligence , pages 1034 { 1038 , 2008 .

[10] D.R. Radev , H. Jing , M. Stys , and D. Tam . Centroid-based summarization of multiple documents . Information Processing & Management , 40 ( 6 ): 919 { 938 , 2004 .

[11] C. Wang , F. Jing , L. Zhang , and H.J. Zhang . Learning query-biased web page summarization . In Proceedings of CIKM , 2007 .

[12] R. Yan , J.Y. Nie , and X. Li . Summarize what you are interested in: An optimization framework for interactive personalized summarization . In In Proceedings of EMNLP , pages 1342 { 1351 , 2011 .