Textmining Screen


The basic screen composition of the patent text mining function is shown below. As a basic usage, you can analyze patents while looking over the plan and time series while rolling the population of patent search results plotted on the left side. The patent search population is drawn so that closer patents are patented as patent-based patent documents by text mining.

Up to now, about 100 to 300 search results (important patents) have to be viewed one by one in order of patent application date, and it is a very troublesome and time-consuming task to overlook and analyze patent search results . Even with about 300 documents at most, it took a lot of trouble when grasping the technology and application trends of other companies.

  • Patent integration will apply text mining technology to patents and support the grasp of the overall picture of the patent search result population by visualizing hundreds of patent documents based on text mining technology. Plot similar documents on the plane and plot each patent document in order of patent application date as depth (Z axis direction). Using the patent text mining function has the following merits.
    • By reviewing similar patents (cluster by clusters), review work will be completed.
    • By arranging similar patents in chronological order (in order of patent application date), by making comparisons and reading, the differences and developments of similar patented subjects become very clear and we can conduct high quality patent investigation.
  • By utilizing patent text mining technology, noise can be easily removed by setting bookmarks for similar documents.

How to use patent text mining function

The patent text mining function can be used by selecting "VIZ" from the patent Search Result Control "ANALYSIS".

Confirm distribution status of patent

Distribution status of patents can be grasped in three dimensions (the depths are plotted in order of filing date) by moving the sliders arranged above and to the left.

Each point corresponds to one patent document, and you can check the outline of patent documents by placing a mouse.

Cluster of patent

Each patent is arranged as closely as the result of calculation by text mining technology "patent with more similar document contents", but since the dimension of the word space of patent calculated by patent integration sometimes extends to tens of thousands of dimensions, Often we can not plot in two dimensions. Patent Integration has the function of clustering (automatic classification) of patents for each patent group according to the similarity (distance) of two patents in the tens of thousands of dimensions space. It is possible to assign color to each cluster freely. For example, it is possible to improve work efficiency by coloring the same color on clusters of almost the same viewpoint.

Also, you can remove points by removing the check box. By eliminating unnecessary points, analysis of the focused cluster becomes easier.

By also considering clustering information, it helps to understand similarity of patent documents calculated by text mining technology in high dimensional space.

The number of clusters is automatically calculated according to the degree of viewpoint included in patent documents. Patent search result population that includes various perspectives is classified into almost many clusters.

Setting flags to patent documents

In the patent text mining function, flags can be set for each cluster, for each patent applicant, and for each patent inventor. Flag settings can be set by operating the combo box listed on the control panel. For example, let's set the "candidate" flag in cluster 4.

The flag has been set.

By setting the flag, it is possible to filter patent documents that set the "candidate" flag from "patent overview" etc. By using this, we can screen (patronize) patent documents with low relevance at the time of patent search, and pick up only relevant patent documents only.

In addition, you can list the grant status of patent classification in patent documents with flags in "patent classification table". By using this function, it is possible to easily identify the patent classification to be searched from the patent preliminary search result, for example. In addition, by registering the specified patent classification in the patent classification bookmark, it can be used when assembling the search expression of this search.

[ Related Article]

Apply processing collectively to multiple patent documents

From each patent document plotted by the text mining function, it is easy to collectively select plural patent documents collectively and set attribute information such as flags, ratings and bookmarks at once, You can also download patents.

Multiple patent documents can be selected at once by " dragging and dropping with the mouse while holding down the shift key ". You can also cancel all selected states by clicking the Release button at the top. The release button also displays the number of selected patent documents.

In selected patent documents the outlines are displayed bold.

Right-clicking at an appropriate position on the screen will display a menu listing the processing that can be collectively applied to the selected patent document. Specifically, you can collect only the selected patent documents at once by bulk browsing, patent browsing, saving them in various formats by downloading patents, setting ratings, bookmarks and flags at once .

To deselect, click the Unlock button located at the top.

Patent search method (Introduction).

Patent search method (Beginner).

Patent Search (Patent Database)

Patent Analysis