Exploring Technology Feature with Patent Analysis

Abstract

The patent literature has documented 90 % of the world’s technological achievements, which are protected by the patent law of each country. But with the increasingly competitive technology, enterprises have started the patent strategy research and attached great importance to patent analysis. The patent analysis uses statistics, data mining, and text mining to convert the information into a competitive intelligence that facilitates corporate decision making and prediction. Thus, the patent analysis has become a corporate weapon for long-term survival and protection of commercial technologies. The patent analysis in the past, compared with the trend analysis, mostly conducted the predictive analysis of a number of keywords and patents through the statistical analysis approach. However, the keywords found were limited to the already mature technology and could not locate the implicit emerging terms, so the patent analysis in the past could only find the words of obvious importance, but fail to find the emerging words that are unobvious yet will have a major impact on future technologies. Therefore, how to find these words of a low-frequency nature to make prediction of the correct trend is an important research topic. This study used the Chinese word segmentation system to find the words of the patent documents and extracted the words according to the probability model of the Cross Collection Mixture Model. This model targets the words under changes in the time series. The background model and the common theme in the model will eliminate frequent words without the meaning of identification and collect words persistently appearing across time. This method can quickly screen enormous volumes of patent documents, extract from the patent summary emerging words of a low-frequency nature, successfully filter out the fashion words, and accurately detect the future trends of emerging technologies from the patent documents.