Research of Technical Development Trend and Hot Points Based on Text Mining

Many methods have been developed to recognize those progresses of technologies,and one of them is to analyze patent information.However,current analysis methods for patent documents have some drawbacks.For example,previous researchers mainly paid attention to the structured item in patent documents such as patent number,filing date other than the unstructured item such as claims,abstracts and so on.However,many methods were not full and lots of valuable information can not be gained. Therefore,we proposed a new method based on text mining to handle not only structured items but also unstructured items in this paper.With collected keywords of a target technology field, we represented patent documents in vector space model and cluster patent documents by the kohonen´s self-organising neural network algorithm. With the clustering results,we formed a semantic network of keywords without respect of filing dates.And a technical development trend graph was built up by rearranging each keyword node of the semantic network according to its earliest filing date and frequency in patent documents. Then we calculateed the value of keywords indices such as the relative growth and relative share and visualize the analysis results to get the technical hot points graph. Our approach contributes to considering both structured and unstructured items of a patent document.Besides,ours visualizes a clear overview of technology information in a more comprehensible way.And as a result of those contributions,it enables us to understand advances of emerging technologies and technical hot points at present,and it helps us to have an insight to the technology field,thereby to avoid unnecessary investments.