trieval- Information Filtering General Terms: Design, Human Factors

Children face several challenges when using information ac-cess systems. These include formulating queries, judging the relevance of documents, and focusing attention on in-terface cues, such as query suggestions, while typing queries [3]. It has also been shown that children want a personalised Web experience and prefer content presented to them that matches their long-term entertainment and education needs [2]. To this end, we have developed an interaction-based information filtering system to address these challenges. The system, a prototype developed within the PuppyIR project [1], is initialised by supplying a list of manually se-lected and vetted syndication feeds. We use sources such as the BBC and other reputable providers to overcome the problem of content moderation1. Interaction with the sys-tem facilitates topic discovery over time; however, a default set of topics are defined to deal with the cold-start problem. Feeds are periodically fetched and split into individual feed entries, which are checked against existing content to ensure they are unique. Finally, documents are filtered for a topic using the Okapi BM25 scoring function with the topic defi-nition acting as a query. Filtered documents are presented in reverse chronological order and the document-query score is used to create visual cues in the presentation of results. The system interface is shown in Figure 1. The set of default topics are shown above the set of discovered topics (1a). In this example, the Science topic has been selected (1b) and the list of documents titles filtered for this topic are presented to the child. The amount of space used in presenting a title indicates the relevance of a document, as shown by comparing a somewhat relevant document (2a) and a strongly relevant document (2b). Children are also able to manually define a new topic (3), if needed, and per-sonalise the interface (4), such as the customizing the title, colour scheme and style