Filtering systems and search engines have traditionally been used in a one-shot mode-the user types a query, a ranking algorithm returns the results, and the interaction ends. However, the input query is at best an imprecise description of the user's information need and that we must engage the user in a dialog to encourage an evolving understanding of what the user was looking for. In this work, a computational linguistics approach for interactive Web-based dialogue interactions aiming at intelligent web search and filtering is proposed. The model focuses on the user's requests by automatically generating language-driven interactions which take into account the context, user's feedback and the initial web search's results. The different components for natural-language processing in the context of dialogue discourse interactions are described. The main results of a working prototype aiming to decrease both the number of conversational turns and the information overload are finally discussed.