Conversational search offers the prospect of improved user experience in information seeking via agent support. However, it is not clear how searchers will respond to this mode of engagement in comparison to a conventional user-driven search interface, such as those found in a standard web search engine. We describe a laboratory-based study directly comparing user behaviour for a prototype agent-mediated multiview conversational search interface (MCSI) which extends the functionality of a conventional search interface (CSI) with that of an equivalent CSI. User reaction and search outcomes of the two interfaces are compared for a set of scenario-based search tasks using implicit evaluation with two analysis methods: workload-related factors (NASA Load Task) and psychometric evaluation. Our investigation shows the MCSI to be more interactive and engaging, with users claiming to have a better search experience in contrast to the corresponding CSI.
Impact-ordered index organizations are suited to score-at-a-time query evaluation strategies. A key advantage of score-at-a-time processing is that query latency can be tightly controlled, leading to lower tail latency and less latency variance overall. While score-at-a-time evaluation strategies have been explored in the literature, there is currently only one notable system that promotes impact-ordered indexing and efficient score-at-a-time query processing. In this paper, we propose an alternative implementation of score-at-a-time retrieval over impact-ordered indexes in the Rust programming language. We detail the efficiency-effectiveness characteristics of our implementation through a range of experiments on two test collections. Our results demonstrate the efficiency of our proposed model in terms of both single-threaded latency, and multi-threaded throughput capability. We make our system publicly available to benefit the community and to promote further research in efficient impact-ordered query processing.
In monolingual retrieval, sparse representations learned atop BERT-style models offer a complementary approach to the unsupervised BM25 model. Inspired by this line of work, we explore adapting such models to the Cross-Language Information Retrieval (CLIR) setting, in which queries and documents are in different languages. The lack of lexical match between queries and documents inhibits a naive replication of these monolingual models for CLIR. We propose SPLADE-X, a cross-language expansion model for CLIR that performs complementary to a strong PSQ baseline. We further identify the challenges in developing such models, make connections to existing CLIR models and highlight future directions that pave the way for learning sparse representations feasible for CLIR.
The BASE prototype aims to improve user awareness of biases in search engine results. It utilises existing resources and NLP tools to identify biases in news articles. It incorporates bias visualisation features to inform users of biases in each news article and at the search results level. It also incorporates results reranking features to allow users to retrieve different sets of results based on their search preferences. Preliminary evaluation results suggest the prototype achieves a positive usability score (64.3 out of 100) and has a potential for increasing user awareness of biases, with the reranking features rated more useful than the bias visualisation features.
