Combining and selecting characteristics of information use

Abstract

In this paper we report on a series of experiments designed to investigate the combination of term and document weighting functions in Information Retrieval. We describe a series of weighting functions, each of which is based on how information is used within documents and collections, and use these weighting functions in two types of experiments: one based on combination of evidence for ad-hoc retrieval, the other based on selective combination of evidence within a relevance feedback situation. We discuss the difficulties involved in predicting good combinations of evidence for ad-hoc retrieval, and suggest the factors that may lead to the success or failure of combination. We also demonstrate how, in a relevance feedback situation, the relevance assessments can provide a good indication of how evidence should be selected for query term weighting. The use of relevance information to guide the combination process is shown to reduce the variability inherent in combination of evidence.

Personalisation and recommender systems in digital libraries

“…As illustrated in Figure 1, the personalisation can be based on different types of characteristics such as characteristics of: a person as both an individual or member of a group (e.g., knowledge or motivitation); the resources or information or documents (e.g., genre of the materials, age or authenticity; and/or perceived outcome (e.g., novelty or accuracy), all of which are related to the media or channel used (e.g., PDA, versus a cell phone or computer), the task that is being performed, and the environment in which the user is immersed, namely the context, 8 , 9 . 44 This list of characteristics emphasizes short-term personalisation and is not comprehensive, but it highlights the relationships between the most important components -namely people, resources and perceived outcomes -and serves as a guide to illustrate the rich types of data that are available and need to be manipulated to personalise and/or recommend.…”

Incorporating user search behavior into relevance feedback

“…The control system in this experiment only performed initial retrievals; there was no relevance feedback component of this system. The basic retrieval algorithm followed the approach given in (Ruthven, Lalmas, & Van Rijsbergen, 2001b). This assigns each term in the collection a set of weights.…”

“…The retrieval score of a document is given by the sum of all the term weights of the query terms contained within the documents. This approach generally gives better results than the more standard tf * idf approaches (Ruthven, Lalmas, & Van Rijsbergen, 2001b).…”

“…This method provides a conservative change to the query: as few expansion terms as necessary are added to the query and the number of terms added is relative to the number of relevant documents found. This method of query expansion was shown to be generally better than adding a fixed number of expansion terms to each query (Ruthven, Lalmas, & Van Rijsbergen, 2001b).…”

Dempster-Shafer Theory for a Query-Biased Combination of Evidence on the Web

“…For example, idf (Spärck Jones, 1972) and its extensions (Aizawa, 2000;Wong and Yao, 1992;Rölleke, 2003) weight terms according to the number of documents they appear in. Moreover, Ruthven et al (2002) define the specificity of a document as the sum of the idf of its terms, divided by the length of the document. From a different perspective, Cronen-Townsend et al (2002), instead of defining a measure of specificity, they model the clarity of a query as the divergence of the query language model from the collection language model.…”