Finding well-known researchers is not difficult. Publication counts, patent counts, and citation counts reveal them fairly well. The hard part is spotting young talent early, people with limited track records who are about to grow. This post walks through a paper that proposes a method for finding such talent using patent data.
The paper is:
- Title: Inventor profile mining approach for prospective human resource scouting
- Journal: Journal of Informetrics (2021)
- DOI: 10.1016/j.joi.2020.101103
Motivation: Existing Methods Were Designed to Find People Who Have Already Made It
For a technology-based firm (TBF), finding young, capable people with knowledge in the firm’s technology fields is important. This task strongly affects the firm’s performance and sustained growth. Earlier studies proposed methods for identifying or recommending key researchers. However, few studies focused on finding young, capable people who fit a firm’s technology strategy, that is, “prospective talent.”
The authors point out three gaps in existing approaches:
- Data: Many studies relied only on academic or web data to evaluate talent. Yet corporate R&D staff usually disclose their R&D results through patents.
- Youth: Youth is hard to express through indicators such as the degree of performance improvement or potential.
- Performance indicators: Citation counts and the h-index (an indicator that reflects both the number of papers and the number of citations) do not capture well the performance of young people with little inventive experience.
In short, the authors’ concern is that existing indicators alone do not reveal the performance of young people with little inventive experience. This paper aims to solve that problem.
Process: Collect Patents, Build Profiles, and Organize Them into Portfolio Maps
The method has three steps:
- Collect patent data related to the target firm and preprocess the patents of candidate inventors.
- Identify each inventor’s technology fields and measure inventive performance and career.
- Build a performance-career portfolio map for each inventive field and identify prospective inventors who fit the target firm’s technology development direction.
The Profile Has Two Parts
The inventor profile built in the second step consists of two parts:
- Inventive fields: These show the technology fields in which an inventor has gained R&D experience.
- Inventive performance and career: These show how capable and how young an inventor is.
Inventive fields are extracted using topic modeling based on Latent Dirichlet Allocation (LDA). Topic modeling is a technique that finds groups of words that frequently appear together in documents and organizes them into topics. Inventive performance is measured with the AW-index, and career is measured by the year of the first patent application.
The reason for using the first application year is clear. The goal is to find people with short inventive careers, not people of a young chronological age. The authors explain that this approach can identify people who are likely to be young from the perspective of inventive career. The AW-index is used to find prospective inventors who achieved strong performance despite limited inventive experience.
Case: Candidate Inventors in Recognition Technology
In Section 4 of the paper, the authors conducted a case study for a specific target firm. To identify candidate inventors’ fields, they used a set of patents containing the patent classification subclass G06K. Some of the fields produced by topic modeling were:
- Decoding and recognition of characters and objects (optical character recognition, object detection, etc.)
- Identification of object and attribute information in printed matter, image, and fingerprint data
- Location information sharing and transmission systems
- Pattern recognition of sensor-based optical data
A noteworthy point is the data used to measure performance and career. The fields were defined using the G06K patent set. However, patents in this set alone can hardly represent an inventor’s full performance and career. Therefore, performance and career were measured using all patents registered by each candidate inventor. Narrowing down the field and fully evaluating the person were handled separately.
Validation
Finally, the authors built performance-career portfolio maps by inventive field. According to the abstract, they conducted a case study and statistical validation and state that these showed the method can identify prospective inventors. This post does not cover the specific validation procedures or result figures.
Significance: Turning Talent Search into a Quantitative Screening Tool
The authors define the prospective inventors found by this method as people who have the technical knowledge the firm needs and can therefore contribute immediately to the firm’s technology strategy. Prospective inventors are also defined as young, capable people who can grow into established inventors and contribute to the firm’s sustained growth. These are definitions set by the paper and expectations held by the authors. The paper does not prove that people found by this method actually perform well after being hired.
Value for practitioners is as follows. Because this method screens prospective talent quantitatively, it can serve as a talent scouting tool. The authors also state that they hope technology-based firms will use it as a human resource management tool. In my view, it has one more advantage. Instead of relying only on HR staff’s intuition or personal referrals, it lets firms first draw up a list of candidates who fit the company’s technology direction.
Value for researchers also exists. Patent data make it possible to evaluate both a person’s R&D experience and performance. The method provides a framework that extends research on identifying key researchers from “people who have already achieved results” to “people who will achieve results.” There is also room to experiment by swapping in different indicators for field identification, performance measurement, and career measurement.
Points to Consider When Applying the Method
Before applying the method in practice, however, a few things should be checked. The following is my own review based on the paper’s design.
- Talent that does not file patents remains invisible. People who work mainly with software or trade secrets, and people who work in university labs with a focus on papers, may be missed.
- The first application year is not chronological age. Experienced professionals who started inventing late may be classified as “young” inventors. The authors also limit their explanation of this indicator to youth from the perspective of inventive career.
- The list is only a starting point. Even candidates who stand out on the portfolio map need separate checks on their willingness to change jobs, team fit, and standards for handling personal information.
Wrap-Up: Try It with Your Company’s Classification Code
Pick one patent classification code in a technology field you care about. From the registered patents under that code, make a table of each inventor’s first application year and number of patents. Simply looking at how many names have short careers but notable patent records lets you see the paper’s concern for yourself.
Business Intelligence Lab introduces technology management research that uses patent and technology data. You can continue reading other paper reviews on our blog.