Text mining approaches applied to patents: A scoping review protocol

English

A	B	C
Review stage	Started	Completed
Preliminary searches	Yes	Yes
Piloting of the study selection process	Yes	Yes
Formal screening of search results against eligibility criteria	Yes	Yes
Data extraction	No	No
Data analysis	No	No

This review is a part of a Ph.D. research project approved by the ethical committee of Kerman University of Medical Sciences, No. 40100010, which will be carried out with the financial support of the Vice Chancellor for Research and Technology of this university. The funding source had no involvement in the study process.

Review question (s)

RQ1. Which text mining techniques are frequently used by researchers in mining patent?

RQ2. Which data sources are the most often used for text mining in patents?

RQ3. Which metadata (Claims, Abstract, Title or description) are frequently used for text mining?

RQ4. In which subject areas is text mining used more in patents?

RQ5. What is the most preferred sample size selected by text mining researchers when applying text mining techniques to patents?

search keywords/terms:

(text* AND (analy* OR mining OR categorization OR classification OR cluster* OR extract* OR preprocessing OR processing OR transformation)), (data AND mining), "document classification", "document cluster*", "document summarization", "machine learning", "keyword extraction", "keyword discovery", "keyword retrieval", (information OR knowledge) AND (extract* OR discovery OR retrieval), "Latent Dirichlet Al", LDA, "Latent Semantic Analysis", LSA, "Natural Language Processing", NLP, "content analysis", "topic extraction", "topic model*", "unstructured text", "unsupervised learning", "Vector Space Model", "VSM", "support vector machines", "naive bayes classifier", "association rules", "k-nearest neighbor", "neural networks" OR "decision trees" AND (patent OR patents)

"patent analy", "patent mining", "patent cluster", "patent map*", "patent roadmap*", "patent network", "patent visualization", "patent** visualisation ", patentometric, "patent classification", "patent retrieval"**

The literature search will not be limited by year of publication or geographic area and the language will be limited to English.

URL to search strategy

We will upload this protocol on Open Sciences Framework (OSF)

We give permission for this file to be made publicly available:

Yes

Condition or domain being studied

Studies that include text mining on unstructured data on patent (such as title, abstract, description, claim).

Participants/ Population

The English Original/Conference papers published that meet the eligibility criteria of this review will be included.

Intervention (s), Exposure(s)

This scoping review does not have Intervention (s), Exposure(s) group.

Comparator (s)/ control

This scoping review does not have Comparator (s)/ control group.

Types of study to be included

The text mining techniques can be utilized to extract the information from structured or unstructured data. In this study our focus will be on unstructured text of patent. The following inclusion and exclusion criteria will apply:

Inclusion criteria:

-Studies that include text mining on unstructured data on patent (such as title, abstract, description, claim);

-peer-reviewed papers published in selected databases and in English;

-publication outlets (Original articles and conference proceedings);

-Full access to the document.

exclusion criteria:

-all publications that do not meet the inclusion criteria;

-Papers which did mining on structured data on patent;

-non-English results will be removed during the review process;

-Articles that focus on just citation analysis of patents;

-Secondary and tertiary studies, such as reviews, meta-analyses and surveys will be drawn;

-Editorial, meeting abstract, reviews, book reviews, books, book chapters and cover letters, and commentaries;

-duplicate publications and retracted publications.

Context

Identification of studies that uses text mining methods on unstructured data on patent (such as title, abstract, description, claim (

Primary outcome(s)

This study will map the current state of different text mining approaches in patents.This study will be of value to policy makers and researchers by allowing them to find the latest efforts to patent text mining

Secondary outcomes

The secondary outcomes will be identifying the different text mining approaches, the most used data sources, metadata and subject areas in different application areas of the patents.

Data extraction (selection and coding)

After running search syntax in each database, the results of all the search will be exported to EndNote 20. Duplicate papers will be removed. The remaining papers will be imported to the Rayyan, for inclusion in review. Two independent research members will screen titles and abstracts of all papers against the inclusion and exclusion criteria. Subsequently the full-text of all potentially relevant papers will be assessed independently by two reviewers. In the case of conflicts, they discussed and then consulted the third author to reach a consensus.

Key findings relevant to the review will be charted from the included studies using a data extraction tool developed in Excel software by the members of the review team. The following data will be extracted: title, author, publication year, Journal, techniques, tools or applications of text mining, data sources, patent metadata (Claims, Abstract, Title or description), subject area, sample size and the purpose of the study. Two review team members will extract data independently and discrepancies will be solved by consulting with a third expert.

The results of the search and the study inclusion process will be reported following the principles of the PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analysis extension for Scoping Reviews)

Strategy for data synthesis

The extracted data will be categorized based on types of different techniques of patent text mining, patent data sources, and key features like comparison of the frequency of use of different patent metadata, identifying the most used subject areas in this topicand sample size.

Analysis of subgroups or subsets

There will be no analysis of subgroups or subsets.

type and method of review

Scoping review

Country

Iran

Current review status

On-going

any additional information

Appendix 1: Search syntax for this Scoping review in WoS:

A	B
	Search syntax
#1	(TS=(( ( text* AND ( analy* OR mining OR categorization OR classification OR cluster* OR extract* OR preprocessing OR processing OR transformation ) ) OR ( data AND mining ) OR "document classification" OR "document cluster*" OR "document summarization" OR "machine learning" OR "keyword extraction" OR "keyword discovery" OR "keyword retrieval" OR ( ( information OR knowledge ) AND ( extract* OR discovery OR retrieval ) ) OR "Latent Dirichlet Al" OR LDA OR "Latent Semantic Analysis" OR LSA OR "Natural Language Processing" OR NLP OR "content analysis" OR "topic extraction" OR "topic model*" OR "unstructured text" OR "unsupervised learning" OR "Vector Space Model" OR "VSM" OR "support vector machines" OR "naive bayes classifier" OR "association rules" OR "k-nearest neighbor" OR "neural networks" OR "decision trees" ) )) AND TS=(patent OR patents )
#2	TS=(( "patent analy*" OR "patent mining" OR "patent cluster*" OR "patent map*" OR "patent roadmap*" OR "patent network" OR "patent visualization" OR "patent visualisation"  OR patentometric* OR "patent classification*" OR "patent retrieval" ))
#1 OR #2	Limited to: Article, Conference paper

Details of final report/publication(s).

It will be published in a peer-reviewed journal.