Sentiment Analysis in the Bio-Medical Domain by Ranjan Satapathy Erik Cambria & Amir Hussain

Author:Ranjan Satapathy, Erik Cambria & Amir Hussain
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

3.2.3 AffectiveSpace

The best way to solve a problem is to already know a solution for it. However, if we have to face a problem we have never met before, we need to use our intuition. Intuition can be explained as the process of making analogies between the current problem and the ones solved in the past to find a suitable solution. Marvin Minsky attributes this property to the so called ‘difference-engines’ [77]. This particular kind of agents operates by recognizing differences between the current state and the desired state, and acting to reduce each difference by invoking K-lines that turn on suitable solution methods.

This kind of thinking is maybe the essence of our supreme intelligence since in everyday life no two situations are ever the same and have to perform this action continuously. To emulate such a process, AffectiveSpace7 [19], a novel affective commonsense knowledge visualization and analysis system, is used. The human mind constructs intelligible meanings by continuously compressing over vital relations [41]. The compression principles aim to transform diffuse and distended conceptual structures to more focused versions so as to become more congenial for human understanding. To this end, principal component analysis (PCA) has been applied on the matrix representation of AffectNet.

In particular, truncated singular value decomposition (TSVD) has been preferred to other dimensionality reduction techniques for its simplicity, relatively low computational cost, and compactness. TSVD, in fact, is particularly suitable for measuring the cross-correlations between affective commonsense concepts as it uses an orthogonal transformation to convert the set of possibly correlated commonsense features associated with each concept into a set of values of uncorrelated variables (the principal components of the SVD). By using Lanczos’ method [60], moreover, the generalization process is relatively fast (a few seconds), despite the size and the sparseness of AffectNet. The objective of such compression is to allow many details in the blend of ConceptNet and WNA to be removed such that the blend only consists of a few essential features that represent the global picture. Applying TSVD on AffectNet, in fact, causes it to describe other features that could apply to known affective concepts by analogy: if a concept in the matrix has no value specified for a feature owned by many similar concepts, then by analogy the concept is likely to have that feature as well. In other words, concepts and features that point in similar directions and, therefore, have high dot products, are good candidates for analogies.

A pioneering work on understanding and visualizing the affective information associated with natural language text was conducted by Osgood et al. [84]. Osgood used multi-dimensional scaling (MDS) to create visualizations of affective words based on similarity ratings of the words provided to subjects from different cultures. Words can be thought of as points in a multi-dimensional space and the similarity ratings represent the distances between these words. MDS projects these distances to points in a smaller dimensional space (usually two or three dimensions). Similarly, AffectiveSpace aims to grasp the semantic and affective similarity between different concepts by plotting them into a multi-dimensional vector space [24].

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.