Improving word similarity computation accuracy by multiple parameter optimization based on ontology knowledge

Word similarity computation is one of the most fundamental areas of research in semantic information processing. Prior studies on Chinese word similarity computation have mostly adopted rule-based methods. Some studies have been conducted on English word similarity computation using the notable knowledge base WordNet. English word similarity computation methods cannot be used directly for word similarity computation. Therefore, we find a ontology knowledge base whose hierarchical structure is similar to WordNet. With the help of it, we develop an improved Chinese word similarity computation method, therein incorporating the common depth, depth parameter, depth adjustment parameter, concept relation parameter, density parameter and differential value into the Chinese word similarity computation process. First, we perform an in-depth analysis on the merits and disadvantages of existing word semantic similarity computation approaches; then, we investigate the effect of several factors on the word semantic similarity computation. Finally, we utilize the hierarchical tree structure of the ontology knowledge base to improve the word similarity computation accuracy. The experimental results show that our proposed method outperforms state-of-the-art methods. Network public opinion is the mapping of social public opinion on the Internet. By using the means of similarity calculation, a platform of online public opinion with prediction and early warning can be built to quickly find the hinge point of public opinion, which provides rich data support for the management.