High secure corporate web search engine

ABSTARCT :

Internet has become the biggest repository of information in the world. It can be considered as a global library where variety of information in different languages and formats is stored in digital form. The volume of information on web is enormous and it has become near to impossible to estimate its size. Because of its size and storing mechanism, finding relevant and precise information has become a difficult task. For searching information from this vast repository, we use search engines. There are thousands of search engines available on internet. For example, if you visit http://www.thesearchen ginelist.com/, you will find a classified list of search engines. This list is category-wise and includes all-purpose search engines in various fields like accounting, blogs, books, legal, medical, etc.

EXISTING SYSTEM :

? A search engine for the Web of Things has to support searching for structured and rapidly changing content, which is a key challenge given that existing Web search engines are based on the assumption that most Web content changes slowly, such that it is sufficient to update an index at a frequency of days or weeks. ? We discover so-called periodic patterns in the time window using a variant of an existing algorithm. ? This allows a seamless integration into the existing Web infrastructure, making it possible to utilize existing applications and services with Web-enabled things. ? In contrast to existing Web search engines, such a real-world search engine has to support searching for rapidly changing state information generated by sensors.

DISADVANTAGE :

? The key challenge that needs to be addressed in constructing a search engine for the Web of Things is the anticipated huge size and extreme dynamics of the search space. ? Analysis indicates that this is an implementation-specific problem caused by memory management issues in the Java virtual machine. ? When a user issues a search request (in this case “room ifw occupancy : empty”), the resolver is in charge of handling its execution. ? For this, it will first reduce the result set to entities which match the static part of the search term (“room ifw”) and feature the requested sensor type(s) (“occupancy”), by querying the index.

PROPOSED SYSTEM :

• We have proposed an approach called sensor ranking, which for a given query computes the probability that a sensor produces the sought output at the time of the query by using indexed prediction models. • Sensors are then pulled in decreasing order of probability until enough matches have been found, thus spending effort first where it counts. • We also exploit the fact that users of search engines are typically not interested in all results, since these are typically far too many to check manually. • We proposed the augmentation of the existing Web with representations of these real-world entities – thus offering online and real-time Web access to the state of the real world.

ADVANTAGE :

? The former tries to optimize the performance of the search engine, while the latter tries to optimize user satisfaction. ? Inaccurate prediction models will, however, degrade the performance of the search engine. ? To sum up, prediction models are periodically created by a sensor gateway, based on a set of recent sensor states, periodically indexed by Dyser, at a rate of days to weeks, and evaluated by Dyser during the resolution of a search request. ? We evaluated the performance of our prototypical search engine in terms of the communication overhead and latency. ? To obtain repeatable results, we used a realistic data set that is replayed by the sensor gateway instead of real sensors.