Screening method for personalized scientific research literature

Abstract

The invention discloses a screening method for personalized scientific research literature. According to the screening method, an algorithm based on vector cosine value comparison is adopted in combination with literature content in a designated scientific research field and relevant scientific literature in the personal research field to obtain the scientific research literature meeting personalized demands finally. Compared with traditional manual literature search, through the screening method, a large amount of literature reading time is saved, scientific research efficiency is substantially improved, the scientific research literature is closer to the user research field compared with scientific research literature found through a manual search method traditionally, and resource waste caused by repeated research is avoided.

Description

technical field

[0001] The invention relates to the field of screening and retrieval methods for scientific research documents, in particular to a screening method for personalized scientific research documents. Background technique

[0002] Scientific research literature is the main carrier of scientific research achievements at present. Under the background of continuous innovation of experimental methods and technological platforms, the number of scientific research literature in various research fields has exploded. When researchers look at a specific research field, they often spend a lot of time screening documents. At the same time, because most of the scientific research documents are in English and the English reading ability of domestic researchers is generally not high, so the reading and searching of scientific research documents is a key issue in our country. A major bottleneck in the field of scientific research.

[0003] Taking lung cancer gene research as an...

Examples

Embodiment 1

[0032] Embodiment 1: as figure 1 Said, a screening method for personalized scientific research literature, said screening method comprises the following steps:

[0033] 1) Collect 3 keywords in specific fields by researchers, input these keywords on the database and get a large number of scientific articles, then output the title, author and abstract information into a summarized text file, and ask the researchers to provide Read 10 preferred literature in this particular field.

[0034] 2) Remove punctuation marks from each article in the summarized text file, separate the articles into individual words, then calculate the TF-IDF value for all words, and establish a k-dimensional vector representing each article;

[0035] by and Represent the vectors of all articles and user preference references, w is the TF-IDF value of each word and k is the number of words;

[0036] word K i in literature d j Term frequency TF in i defined as ;

[0037] word K i Inverse Docum...

Embodiment 2

[0044] Embodiment 2: as figure 1 and 2 Shown:

[0045] In 2016, there are more than 10,000 related literatures in three distinct research areas: new application of old drugs, pancreatic cancer gene targets and metabolic mechanism of fatty liver. According to the traditional manual search method, it is impossible to complete a comprehensive reading of all literatures Yes, assuming that 5% of the documents are randomly selected for reading, it will take 3-6 months (estimated according to the difficulty of different documents).

[0046] However, searching according to the method of the present invention, we spent about 1 day to complete the comprehensive reading of the literature and recommend personalized scientific research literature. In addition, after perfecting the training data set for 4 rounds, we sampled and read the recommended documents according to the traditional manual reading method for manual judgment, and found that the accuracy rate of the recommended document...

Embodiment 3

[0047] Embodiment 3: as figure 1 and 3 Shown:

[0048] In 2016, there were more than 30,000 related literatures in three distinct research areas: pancreatic cancer new drug trials, hepatotoxic inflammasomes, and cholestasis mechanism. According to the traditional manual search method, it is impossible to complete a comprehensive reading of all literatures. Yes, assuming that 5% of the documents are randomly selected for reading, it will take 6-9 months (estimated according to the difficulty of different documents).

[0049] However, searching according to the method of the present invention, we spent about 6 days to complete the comprehensive reading of the literature and recommend personalized scientific research literature. In addition, after perfecting the training data set for 4 rounds, we sampled and read the recommended documents according to the traditional manual reading method for manual judgment, and found that the accuracy rate of the recommended documents meeting...

technical field

[0001] The invention relates to the field of screening and retrieval methods for scientific research documents, in particular to a screening method for personalized scientific research documents. Background technique

[0002] Scientific research literature is the main carrier of scientific research achievements at present. Under the background of continuous innovation of experimental methods and technological platforms, the number of scientific research literature in various research fields has exploded. When researchers look at a specific research field, they often spend a lot of time screening documents. At the same time, because most of the scientific research documents are in English and the English reading ability of domestic researchers is generally not high, so the reading and searching of scientific research documents is a key issue in our country. A major bottleneck in the field of scientific research.

[0003] Taking lung cancer gene research as an...

Examples

Embodiment 1

[0032] Embodiment 1: as figure 1 Said, a screening method for personalized scientific research literature, said screening method comprises the following steps:

[0033] 1) Collect 3 keywords in specific fields by researchers, input these keywords on the database and get a large number of scientific articles, then output the title, author and abstract information into a summarized text file, and ask the researchers to provide Read 10 preferred literature in this particular field.

[0034] 2) Remove punctuation marks from each article in the summarized text file, separate the articles into individual words, then calculate the TF-IDF value for all words, and establish a k-dimensional vector representing each article;

[0035] by and Represent the vectors of all articles and user preference references, w is the TF-IDF value of each word and k is the number of words;

[0036] word K i in literature d j Term frequency TF in i defined as ;

[0037] word K i Inverse Docum...

Embodiment 2

[0044] Embodiment 2: as figure 1 and 2 Shown:

[0045] In 2016, there are more than 10,000 related literatures in three distinct research areas: new application of old drugs, pancreatic cancer gene targets and metabolic mechanism of fatty liver. According to the traditional manual search method, it is impossible to complete a comprehensive reading of all literatures Yes, assuming that 5% of the documents are randomly selected for reading, it will take 3-6 months (estimated according to the difficulty of different documents).

[0046] However, searching according to the method of the present invention, we spent about 1 day to complete the comprehensive reading of the literature and recommend personalized scientific research literature. In addition, after perfecting the training data set for 4 rounds, we sampled and read the recommended documents according to the traditional manual reading method for manual judgment, and found that the accuracy rate of the recommended document...

Embodiment 3

[0047] Embodiment 3: as figure 1 and 3 Shown:

[0048] In 2016, there were more than 30,000 related literatures in three distinct research areas: pancreatic cancer new drug trials, hepatotoxic inflammasomes, and cholestasis mechanism. According to the traditional manual search method, it is impossible to complete a comprehensive reading of all literatures. Yes, assuming that 5% of the documents are randomly selected for reading, it will take 6-9 months (estimated according to the difficulty of different documents).

[0049] However, searching according to the method of the present invention, we spent about 6 days to complete the comprehensive reading of the literature and recommend personalized scientific research literature. In addition, after perfecting the training data set for 4 rounds, we sampled and read the recommended documents according to the traditional manual reading method for manual judgment, and found that the accuracy rate of the recommended documents meeting...

Images