XNUMX. XNUMX.Introduction

　In a patent search, we conceptualize elements from the contents of the search and create a search population by combining patent classifications such as F-terms and keywords.However, in the case of searchs that include elements that are difficult to verbalize, such as "chemical structures," it may be difficult to conceptualize the elements.I will introduce some of the research methods that I have learned so far regarding research that includes elements that are difficult to verbalize such as "chemical structural formulas".

XNUMX.Population creation

XNUMX Creating a population by combining patent classifications and keywords

　First of all, the following methods are conceivable when creating a population that includes elements that are difficult to verbalize, such as "chemical structural formulas," by combining patent classifications and keywords.

① Use concepts other than fields and "chemical structures"
② Use patent classifications and keywords related to elements and functional groups in the "chemical structural formula"

　However, if we create a population by combining patent classifications and keywords for elements that are difficult to verbalize, we find that not only documents that contain the "chemical structural formula" itself, but also a low precision population that includes many documents other than the target of the search. There is a risk of becoming a group.

XNUMX Population creation using structure search

　Therefore, it is very effective to create a population using structure search in the case of a search that includes elements that are difficult to verbalize, such as "chemical structural formulas."Structural search is a method of searching from the "chemical structural formula" by directly writing the "chemical structural formula". You can search for derivatives and collect documents that describe their "chemical structural formula".When I perform a structure search, I use the "CAS STNext (hereafter STN)” to create the study population.

　To briefly explain the search method, first "STN” to search for the structure of the desired “chemical structural formula” using the “REGISTRY file”, which is a database of chemical substances.Next, by crossing over the search results of the structure search in the "REGISTRY file" to the "files in line with the research purpose in CAS FILES (CAplus, etc.)", the target "chemical structural formula" is described (registration) HIT the literature that is written.
※For more information【JAICI Chemical Information Association STN Chemical Substance Search] Please refer to.

XNUMX.Fusion of Populations

　Now, as described above, we have created two populations, a population combining patent classifications and keywords, and a population using structure search. Look at the characteristics of the group.

　First, in the population using structure search, documents that describe (register) the "chemical structural formula" you want to search for are hit, and generally the population has a high matching rate.However, publications that do not indicate specific substances and are described only in Markush format, or in which "chemical structural formulas" are expressed only in language, may be dropped from the population. .

　Next, the population that combines patent classifications and keywords can include the target compound and its superordinate concept compounds in the population, and generally becomes a population with a high recall rate.However, if the patent classification is not assigned to the target publication or the keyword is not used, the target publication may be dropped.

　Then, it would be better to conduct a search in two populations, but there is a risk of duplicating the same literature in each population.In addition, it is more efficient to conduct peer review using a patent database that you are familiar with on a daily basis.

　Therefore, as a method I am doing, first of all,STNDownload the publication number from the results of [Structure Search] usingNext, look up the publication number in your usual patent database.Then, combine the population that combines the patent classifications and keywords created in the patent database that you normally use with the population that uses the structure search to eliminate duplication.After merging the two populations into one population in this way, we will proceed with the search peer review using the patent database that we normally use.

　By proceeding in the manner described above, it is possible to create a single population that approaches the research content from multiple perspectives, such as structural search, patent classification, and keywords.

XNUMX.in conclusion

　[Structure search] is very effective as one of the research methods including elements that are difficult to verbalize such as "chemical structural formula".Then, by combining a population that combines patent classifications and keywords, a population with a higher recall rate can be created and the probability of achieving the research purpose can be increased.In addition, in the fields of chemistry, biotechnology, and medicine, there are factors other than "chemical structural formulas" that are difficult to verbalize, such as proteins and nucleic acid sequences. By combining populations, you will be able to create a population with a higher recall rate.

Research Division Akiba

【reference】

CAS STNext
https://www.jaici.or.jp/stn-ip-protection-suite/cas-stnext/

JAICI Chemical Information Society STN Chemical Substance Search
https://www.jaici.or.jp/application/files/2716/5354/7916/text_chem.pdf