Ranking the Value of Public Posts 22/09/2009
Two patents were awarded to Microsoft Corporation earlier this month covering methods for identifying and categorizing the gobs of disparate information types that float in cyberspace. One similarity in both methods is the generation of feature vectors based partly on the extraction and ranking of keywords in relation to other words from the same post. The first patent U.S. 7,590,612 applies to the blogosphere, and it aims to map blog posts to help with the navigation and searching of blogs. According to the patent, blogs are typically user-defined material which makes the content difficult to classify into distinct categories. Thus, the patent proposes a computational mapping method to make it easier for users to know when there is a change in topic, what topics were being discussed and where to find specific blog posts which interests them. Each blog post is converted to a feature vector to be placed on the map and labeled. By employing various tools, the attributes of each feature vector can be scaled up or down in complexity. The second patent U.S. 7,590,603 applies to online discussion threads, typically those providing customer support assistance. The patent proposes a computer-based classifier which is trained to distinguish when a message is deemed to be a question type message. According to the method described, the classifier is capable of sorting out messages which are relevant to a user query, then finds a suitable answer which had already been given in an earlier message, and display this answer to the user. The classifier is also capable of ranking answer type messages in terms of relevancy to a specific query.
Add Comment | About EntréeEntrée is the blog segment of EnIPma's website, and it shall include from time to time snippets of information related to the world of patents, R&D and related investment activities.  ArchivesJuly 2010 CategoriesAll |
RSS Feed