PeRSSonal’s core functionality evaluation: Enhancing text labeling through personalized summaries

Noun retrieval effect on text summarization and delivery of personalized news articles to the user’s desktop

“…The main scope of the categorization module is to assist the summarization procedure by pre-labeling the article with a category and has proven in [3] to be providing better results, as far as summarization is concerned. Following the IR tasks of the system, personalization algorithms take place.…”

“…Common sense implies that a keyword that has very high frequency for a specific category, should give more weight to the sentence in which it appears, while a keyword that has small or zero frequency for a category could add less to the weight of a sentence. This factor, namely k 3 in [3], is the keyword's ability to represent the category to which the document belongs. More specifically, k 3 can express the positive or negative effect that a keyword's category has on the summarization procedure and is automatically tunable by the core system processes.…”

“…Moreover, when a keyword is tagged as a noun, we consider it significant thus boosting it with some extra weight. Parameters k 1 and k 2 are thoroughly explained in [3], where through experimental procedure we have concluded values for k 1 and k 2 based on statistical results. N derives from the following equation:…”

“…This implementation offers a significant boost to our application over the web interface of PeRSSonal, since it makes it more responsive and usable. For instance, consider the following scenario: a user's profile contains many keywords from the politics category with high preference (as explained in Section 5.4) and from the fetched ''latest" channel's articles, some of them are related with high similarity to this category -more precisely, the similarity check is done upon the keywords, so an article belongs with high similarity to a category when the majority of its keywords belong to this category, see [3] (categorization subsystem) for more details -the article's summary and relativity information are automatically loaded and stored for future use. Fig.…”

“…Following the keyword extraction procedure, text summarization and categorization techniques come. Both of them operate on a keyword basis, working in a bipartite interactive way [3]. Moreover, user profiling serves another significant role in our system allowing a continuous feedback from the users, enhancing in a secondary stage the resulting information.…”