Evaluating subtopic retrieval methods: Clustering versus diversification of search results

Dualism in Topical Relevance

There are several concepts whose interpretation and meaning is defined through their binary opposition with other opposite concepts. To this end, in this paper we elaborate on the idea of leveraging the available antonyms of the original query terms for eventually producing an answer which provides a better overview of the related conceptual and information space. Specifically, we sketch a method in which antonyms are used for producing dual queries, which can in turn be exploited for defining a multi-dimensional topical relevance based on the antonyms. We motivate this direction by providing examples and by conducting a preliminary evaluation that shows its importance to specific users.

Perspective-based search: a new paradigm for bursting the information bubble

Nowadays, individuals heavily rely on search engines for seeking information. The presence of information bubbles (filter bubbles and echo chambers) can threaten the effectiveness of these systems in providing unbiased information and damage healthy civic discourse and open-minded deliberation. In this paper, we propose a new paradigm for search that aims at mitigating the information bubble in the search. The paradigm, which we call perspective-based search (PBS), is based on the intuition that in a fair search the user should not be limited to the results corresponding to a specific perspective of the search topic. Briefly, in PBS, different perspectives of the search topic are identified and presented to the user and the user can select a perspective for the search results. In this paper, we focus on the paradigm itself, why it is an appropriate solution, and how it differs from other solutions. We raise new questions and call for research on the paradigm and on providing solutions for implementing its required components. We do not aim at providing any specific implementation for it, although we provide some hints on implementing it. We also provide a survey of the related concepts and methods and discuss their differences with PBS.

Refining Kea++ automatic keyphrase assignment

Keyphrases facilitate finding the right information in digital sources. Keyphrase assignment is the alignment of documents or text with keyphrases of any standard taxonomy/classification system. Kea++ is an automatic keyphrase assignment tool using a machine learning-based technique. However, it does not effectively exploit the hierarchical relations that exist in its input taxonomy and returns noise in its results. The refinement methodology was designed as a top layer of Kea++ in order to fine tune its results. It was an initial step and focused on a single Computing domain. It was neither validated on multiple domains nor evaluated to determine whether the improvement in the results is significant or not. The aim of this task was to solidify the refinement methodology. The main contributions of this work are (a) to extend the methodology for multiple domains and (b) to statistically verify that the improvement in the Kea++ results is significant.

DIVERSIFYING SEMANTIC ENTITY SEARCH: INDEPENDENT COMPONENT ANALYSIS APPROACH

Structured Web data are increasingly accessed using information retrieval methods and information retrieval increasingly relies on structured background knowledge. As users' searches are often directed towards finding information about entities rather than text documents, a key affordance of semantic search is the ability to retrieve relevant information about entities more precisely by utilizing the rich structured descriptions and background knowledge. Entity search also poses challenges for information retrieval methods. Entity descriptions are often short and conventional search term matching alone can be insufficient. As a consequence, the search engine should be able to increase the recall of the returned results and select a representative set of entities for a user; to diversify search results. This paper presents an approach to diversify entity search by using semantics present and inferred from the initial entity search results. Our approach utilizes ontologies as a source of background knowledge to improve recall of entity retrieval and independent component analysis to detect independent latent components shared by the entities. The search results are then diversified by selecting a representative set of entities based on their membership in the independent components. We demonstrate the performance of our approach through retrieval experiments conducted by using a real-world dataset composed from four entity databases. The results suggest that our approach can significantly improve effectiveness and diversity of entity search.