Information Technology Reference
In-Depth Information
researches are either lack of experiments or only classify patents according to Inventive
Principles but not related to Engineering Parameters. In this paper, we focus on classify-
ing the contradiction problem of the patents based on Engineering Parameters.
There are many challenges of classifying patents based on Engineering Parameters.
First, the amount of data is very limited because that nearly all of the public domain
databases or websites related to patent classification do not contain the information
about TRIZ. The information of TRIZ can only be manually classified by experts.
Second, the distribution of the conflicting Engineering Parameters of patents is imba-
lanced. Third, some Engineering Parameters have very similar concept and definition.
Fourth, some Engineering Parameters related to the documents are not tagged because
they do not cause the main conflict. This means the data are partially labeled or in-
complete because there is no negative instance.
The contributions of this article are summarized as follows. First, this paper pro-
posed an algorithm named Multi-layer Classification Including Verb Consideration
(MCIVC) which is used for classifying patents based on Engineering Parameters.
Second, it considers positive and negative words to improve the performance. Third,
we proposed the algorithm Verb Including Split and Associate Termsets (VISAT) to
find the relative candidate term sets automatically. The experimental results show that
the MCIVC achieved good efficacy of classifying patent contradictions based on the
Engineering Parameters.
The remainder of this paper is organized as follows. In Section 2, we discuss some
patent analysis software and researches. In Section 3, we introduce the proposed
framework of MCIVC and VISAT. Experimental evaluations and discussion are in
Section 4. The conclusion will be described in Section 5.
2
Related Works
Our study focuses on classifying technical contradiction of patent documents based on
conflicting Engineering Parameters. Some computer-aided innovation (CAI) software
for TRIZ was proposed, such as InventionTool [4], Techoptimizer [5], Pro/Innovator
[6], Creax [7] and Goldfire [8]. Most of this type of software only provides the cor-
responding Inventive Principles when contradictions of the problems are already de-
fined. Except the software, the more related research issue is patent analysis. Some
researches making efforts in estimating TRIZ level of invention of patents had been
proposed [9,10]. This type of researches can be used to filter patents which never
solved any technical contradiction. The algorithm proposed by Gaetano Cascini and
Davide Russo [11] constructs the Subject-Action-Object model combining with the
predefined specific patterns and the morphological patterns to extract useful informa-
tion, but it cannot map the extracted features to the certain categories directly. The
most relevant research issues with our research are the patent classification. There
are many researches proposed for automatic patent classification [12,13,14,15,16,17],
but most of them classify patents according to the technical fields, not the TRIZ in-
formation. Recently, there are some researches about classifying patent documents for
Search WWH ::
Custom Search