Continuous Top-k Monitoring on Document Streams

In the age of the real-time web, online news media aggregators, like Google News, and social media sites, like Twitter, strive for effective and efficient filtering of great volumes of information for millions of users. Given the massive diversity and burstiness of information published on the web every minute, filtering, ranking as well as delivering content streams to interested users come to be a challenging task. Furthermore, feedback signals on the content, such as clicks or shares, offer beneficial information on content significance, but also need more complex manipulation techniques for filtering these streams. Scoring functions in this context cogitate both static as well as dynamic ranking factors, such as profile relevance, recency of information and dynamic feedback signals. Present works for the real-time web fail to handle such dynamic scoring functions in an online way as well as rather adjust an approach of iterative execution of snapshot queries. In this work we proposed effectual evaluation techniques of continuous top-k queries over text as well as feedback streams including generalized scoring functions which capture dynamic ranking features. As a first contribution, we generalize state of the art continuous top-k query models, by presenting a common family of nonhomogeneous scoring functions merging query-independent item significance with query-dependent content relevance as well as continuous score decay reflecting information freshness. Our second contribution involves in the definition as well as implementation of efficient in-memory data structures for indexing and evaluating this new family of continuous top-k queries. Our experimentations show that our solution is scalable and outperforms other present state of the art solutions, when restricted to homogeneous functions. Going a step further, in the second portion of this thesis we study the problem of incorporating dynamic feedback signals to the original scoring function in addition to propose a new general real time query evaluation framework with a family of new algorithms for resourcefully processing continuous top-k queries with dynamic feedback scores in a real-time web context.