Browse Prior Art Database

Method for Automatically Reviewing Academic Publications and Patents based on Peer Review to Judge Structure, Background, Originality and Technical Quality

IP.com Disclosure Number: IPCOM000250536D
Publication Date: 2017-Jul-29
Document File: 3 page(s) / 214K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method is disclosed for automatically reviewing academic publications and patents based on peer review to judge structure, background, originality and technical quality.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

1

Method for Automatically Reviewing Academic Publications and Patents based on Peer Review to Judge Structure, Background, Originality and Technical Quality

Technological advancements are usually shared in the form of publications or patent documents. The technical documents (both patents and publications) requires review to assess quality, relevance, originality and technical rigor before the document is accepted for publication. As the growth in number of publications/patents is exponential in recent years, reviewing the same takes huge resources which is a bottleneck for distribution of knowledge. So, there exists a need for automatically reviewing academic publications and patents based on peer review.

Disclosed is a method for automatically reviewing academic publications and patents based on peer review to judge structure, background, originality and technical quality. The method initially makes assessment separately for each category (structure, background, originality and technical quality) in the article for providing recommendations, by comparing the article with already published articles in the state of the art. The recommendations reflect coverage of the article in a relevant category, which are then used by human reviewers as a preliminary review before submission allowing authors to identify and correct gaps in the article/document.

In accordance with the method and system, existing literature is initially processed to obtain a claim database and statistics over the structure of articles in the literature. The obtained module (claim database and statistics) is then applied for all the categories. All claims from the claim database are then automatically extracted and linked for identifying sections of the articles in which the claims are made. The sections of the articles and main claims made in each of the sections are identified by means of Natural Language Processing (NLP) techniques. Thereafter, a main claim index is generated from the pre-processed claim database and article structure statistics stored in a separate database. Additionally, domain of the articles is automatically detected by relying on information from abstract or metadata (if available). The identified claims are then compared with a database of claims obtained from the literature, in order to provide recommendations for the article in each of the four categories.

Figure 1, illustrates processi...