Topological Research On User Interest In Patent Literature And Application In Pushing Voluntarily Micro Services

The patent literature contains rich information of knowledge innovation and is a kind of important knowledge carrier, which has an important reference value.Reading behavior can reflect user interest in a certain.By combining knowledge map and qualitative research methods, this dissertation analyzed the present situation of domestic and foreign researches and found that the research of reading behaviorfocusing on the patent literature is in the blank. In this dissertation, considering the unique structure style and content of patent literature, eye movement experiment was conducted to explore the toplogy of user interest in patent elements. Further more, interest model was established to calculate the user’interestingness, which is used to set position weights of items. The items that users use frequently, the corresponding IPCs of these items, and the position weights are comprehensively considered to find the favorite patent documentsfor the user and provide the notificatation in microservices.The content and innovation of this study includes the following aspects.(1) Analysis of knowledge map of reading behavior interests in China and abroadBased on a statistical analysis of the articles on eye movement study in reading behavior from Web of Science with Cite SpaceⅢ, Hist Cite and online analysis of the database, this dissertation identifies the intellectual base and the hot issues in this field including the distribution of categories, the important countries, institutions, research focus and fronts, author co-citation, document co-citation and journal co-citation.(2) Theoretical model construction of user interest in reading behavior of patent literatureThrough the Tobii T60 XL of eye movement data for processing and analyzing, in order to find the preference differences in each region of interest. and through the observation to the line of sight scanning trajectory analysis to obtain the process of line scan mode, and further to determine the indicators of interest model, including: relative visit duration, relative fixation count, pupil diameter scaling and regression count;by using the improved analytic hierarchy process to design the weight of the index of interest model, provides a general method to construct the user’s patent literature reading interest model.(3) Mining reader’s interest in patent literature and establishing topology for interest representations in elementsWith the Tobii T60 XL eye tracker as a tool, this dissertation attempts to carry out experimental study by obtaining data from the reading of patent literature. The study considered twenty-six participants mainly university teachers and graduate students with normal vision who are actively engaged in scientific research. For the unique layout structure of patent literature, each patent literature was divided into twelve areas of interests(AOIs) or elements, based on the participants’ eye movements behavior in patent reading, presession interview script and post-session retrospective think aloud(RTA) interviews; interestingness of the element calculated by eye tracking metrics of relative visit duration, relative fixation count and pupil diameter scaling is based on AOI’s; obtaining the relation matrix. By analysis of the interrelated interests between AOIs, the topology of elements was constructed with respect to participants’ interest.(4) Construction of patent active recommendation modelBy mining of user interest in patent literature, we construct user micro service recommendation model. Construction of patent literature structure tree model and text preprocessing,using interestingness to assign characteristic words micro interest weight, using the interest topology to assign the positional weight of the feature word,and combined with the length of the feature word to improve the TF-IDF algorithm, computing comprehensive weight of characteristic words,so as to construct a model of patent feature words vector with user interest;then dig the IPC small class corresponding to the characteristic words of interest,defining the patented technology area to be retrieved,and then search the results in a patent data source.Implementing the similarity matching calculation, ranking of the user interest of the patent feature words and the retrieval results,and getting the TOP user’ high interest patent literature.