A semantic network, or frame network, is anetwork threpresents semantic relations between concepts. This is often used as a form of knowledge representation. It is a directed or undirected graph consisting of vertices, which represent concepts, and edges, which represent semantic relations between concepts.IMAGE :
human
^
|
| politician --is--> in party y
| ^
is-a |
| is-a
| |
| |
Ms. X ----is--------> MP ---can---> vote in p.
The hypothesis for educational technologists is that mapping the semantic network of an expert or knowledgeable person onto the structure of a hypertext and then exposing the learner's to it will contribute to the development of the learner's knowledge structuresResearch shows that semantic networks are somewhat transferable: as a result of instruction, learners' knowledge structures more closely resemble the instructor's knowledge structure. So, learners are acquiring two things during instruction:
- isolated knowledge
- knowledge structures that mimic the teacher's knowledge structure
-- Advantages:
-- Easy to implement, understand and use. There are many practical and scalable implementations available. Some of them are mahout (java), gensim
-- The mahout implementation can train on big data sets, provided you have computational resources.
-- Performance: LSA is capable of assuring decent results , much better than plain vector space model. It works well on data-set with diverse topics.
-- It is not sensitive to starting conditions (like neural network) , so consistent.
-- Apply it on new data is easier and faster compared to other methods. The matrix in topics space has to be multiplied with the vector to get the latent vector of a new document.
Disadvantages:
-- Since it is a distributional model, so not an efficient representation, when compared against state-of-the-art methods (say deep neural networks).
-- Representation is dense, so hard to index based on individual dimensions.
-- It is a linear model, so not the best solution to handle non linear dependencies
-- Deciding on the number of topics is based on heuristics and needs some expertise. Conventional solution is to sort the cumulative singular values in descending order and finding a cut .
THE SIMPLE FIGURE:--
THE SIMPLE FIGURE:--
