Ultimately, hybrid neural-symbolic systems concern the use of problem-specific symbolic knowledge within the neurocomputing paradigm. Neural nets tend to be especially good at recognition of patterns in highdimensional quantitative data, among other things. Symbolic systems tend to be good at abstract reasoning and syntax processing, among other things. Both kinds of systems can be good at generalization and analogy, and at credit assignment, but in different contexts and in different ways. What we achieve by interweaving both is faster real-world modelling, combined with meaningful expression of results and links, automated link-building, analytics and reporting.
Apply trainable deep learning architectures based on the use of convolutional neural networks to extract both local contextual features and global contextual features from text. The higher layers in the overall deep architecture will make highly effective use of the extracted context - sensitive features to perform semantic matching between documents and queries, both in the form of text, for Web search applications.
Many current solutions are complicated, consist of several stages and hand-built features, and are too slow to be applied as part of real applications that require such semantic labels, partly because of their use of a syntactic parser. Our ANN-based algorithms instead can do direct mapping from source sentence to semantic tags for a given predicate without the aid of a parser or a chunker. The resulting system obtains accuracies comparable to the current state-of-the-art at a fraction of the computational cost.
Think of an Ontology Driven Agent Based Focused Crawler that attempts to improve the efficiency of queries by introducing semantics in which a keyword is searched. It follows a context-based approach that analyzes content of web page thereby reducing the redundant information and hence deduces the relevant information from a page. Multiagent neural network systems are an alternative effective solution to large scale Web mining. Such a neural network based framework for mining content of semantic web can provide query relevant knowledge using clustering technique. Both methods allow us to provide context based knowledge oriented results to the user.
Neural Network techniques can be useful for tasks like text routing applied to different architectures and different corpora. By combining ANNs and Semantic technologies, better recall and precision results are achieved because the influence of misleading word representations is limited. In addition recurrent neural networks can learn to process messages in a faster and more robust manner. Moreover, neural networks can be used effectively for better and faster ontology alignment.