Semantic multi-dimensional scaling for open-domain sentiment analysis

The ability to understand natural language text is far from being emulated in machines. One of the main hurdles to overcome is that computers lack both the common and common-sense knowledge humans normally acquire during the formative years of their lives. In order to really understand natural language, a machine should be able to grasp such kind of knowledge, rather than merely relying on the valence of keywords and word co-occurrence frequencies. In this work, the largest existing taxonomy of common knowledge is blended with a natural-language-based semantic network of common-sense knowledge, and multi-dimensional scaling is applied on the resulting knowledge base for open-domain opinion mining and sentiment analysis.