Article Preview
TopIntroduction
The current human population is blessed with the ease of information availability on digital platforms. The same blessing turns into a bane of information overwhelm when a user does not know what to search for. A study performed by Lyman et al. (2000) showed that the world's annual output of web content is roughly 1.5 million terabytes. Similar observations can be made on Wikipedia. Wikipedia is the world's largest crowd-sourced encyclopedia. Since its inception, there has been an exponential increase in the number of articles present on Wikipedia (Oecd & OECD, 2010). Users can currently access Wikipedia articles through one of the two methods: First, by browsing on either a subject index or a title index sorted alphabetically, and second, by following hyperlinks embedded within article pages, called WikiLinks (Lüer & Cummins, 2009). These access methods are static since they are subjected to manual editing.
Figure 1. Plot demonstrating expeditious increase in the number of articles in English Wikipedia (Source: PerSummRe_images\wiki_size300)
English Wikipedia alone contains more than 6 million articles, over 3.6 billion words. It has as many words as the 120-volume English-language Encyclopædia Britannica (online), and more words than the enormous 119-volume Spanish-language Enciclopedia universal ilustrada europeo-americana . Figure 1 shows the rapid increase in the size of English Wikipedia. Unfortunately, not all Wikipedia articles are of interest to the user. Presenting the user with a synopsis of the interesting articles can save the user by identifying which articles are most relevant to them.
This can either be a generic summary, which gives an overall sense of an article's content, or a personalized summary, which presents the filtered content as per the user's interest. Several researchers have investigated various approaches to create personalized document summaries (Park, 2008; Y. Liu et al., 2008). These approaches require the sharing of personal information or some complex algorithm to create summaries.
Figure 2. Overall workflow of PreSummRe operations (Source: PerSummRe_images\workflow300)
Eye gaze tracking provides a new dimension to capture a user's ROI (Region Of Interest). But to create summaries using eye gaze, users are required to read the article during run-time. Personalized recommendations are a vital method for information retrieval and content discovery in today's information rich environment. Recommendation, combined with summarization, can allow users to face a huge amount of information to navigate that information efficiently and satisfyingly. In this paper, the authors present a novel approach to recommend personalized summaries for Wikipedia users. The proposed method depends on the collaborative filtering-based recommendation. It extracts information from the past summaries created by the user to identify similar users in-terms of the reading pattern. Then it uses the summaries of these users on the mentioned article to create a recommended summary. Note that this is different from merely creating a new summary using the reader's eye gaze. Here, user behavior has been analyzed based on his/her past reads and then recommend the sentences from the new article, which are of the user's interest.