On-line new event detection and tracking

Self-adaptive user profiles for large-scale data delivery

“…The user profile is initialized to be empty. 3 We then present the documents in our training set to the system along with the corresponding feedback values obtained from the synthetic profile as defined in Section 4.2. After training is completed, the profile algorithm is disabled (i.e., the profiles generated are frozen) and the system is evaluated by having the profile it generated score and rank-order a set of test documents that it has not yet seen, based on their likelihood of relevance.…”

“…There has been significant research on text-based profile construction in information retrieval community (e.g., [5,1,7,3]), especially in the framework of TREC [24]. The main emphasis of TREC, however, has always been on the effectiveness of the participating systems, rather than on their efficiency.…”

LPTA: A Probabilistic Model for Latent Periodic Topic Analysis

“…Event detection and tracking In [1], Allan et al introduced the problems of event detection and tracking within a stream of broadcast news stories. To extract meaningful structure from document streams that arrive continuously over time is a fundamental problem in text mining [13].…”

“…Nowadays with the development of the Web, many text data exist with time information, e.g., news articles associated with their publishing dates, tagged photos annotated with their taken dates in Flickr 1 and published tweets along with their upload times in Twitter 2 . A lot of useful information is embedded in these text data, and it is interesting to discover topics that are periodic and characterize their temporal patterns.…”

Detecting Deviations in Text Collections: An Approach Using Conceptual Graphs

“…The first approach focuses on detecting rare texts in a given collection (Alexandrov et al, 2000). The second allows detecting rare topics in a collection (Feldman and Dagan, 1995;Allan et al, 1998). …”