Challenges in Implementing a University-Based Innovation Search

Engine

Arman Arzani

, Marcus Handte

and Pedro Jos

e Marr

University of Duisburg-Essen, Essen, Germany

Keywords:

Knowledge Transfer, Founding Potential, Researcher Proﬁling, Innovation Identiﬁcation.

Abstract:

In universities, technology transfer plays an important role in the joint development and dissemination of

knowledge as a product that beneﬁts society through innovation. In order to facilitate knowledge transfer,

many universities hire innovation coaches that employ a scouting process to identify faculty members and

students who possess the requisite knowledge, expertise, and potential to establish startups. Since there is

no systematic approach to measure the innovation potential of university members based on their academic

activities, the scouting process is typically subjective and relies heavily on the experience of the innovation

coaches. In this paper, we motivate the need for INSE (INnovation Search Engine) to support innovation

coaches during their search for innovation potential at a university. After discussing the information needs of

the scouting process, we outline a basic system architecture to support it, and we identify a number of research

challenges. Our aim is to motivate vigorous research in this area by illustrating the need for novel, data-driven

approaches towards effective innovation scouting and successful knowledge transfer.

1 INTRODUCTION

Technology transfer is central to the development of

an iconic entrepreneurial university. Academic sci-

ence has become increasingly entrepreneurial, not

only through industry connections for research sup-

port or transfer of technology but also in its inner

dynamic. Many universities are expanding their tra-

ditional roles beyond education and research to in-

clude knowledge transfer, which involves joint devel-

opment and dissemination of knowledge as a product

that beneﬁts society through communication, experi-

ence sharing, building contacts, and innovation net-

works. There are various forms of technology trans-

fer at universities, including the marketing of patents

through licensing agreements, the formation of joint

ventures with industrial partners as well as the cre-

ation of startups that aim to commercialize an inno-

vative research result (Bliznets et al., 2018). While

all these forms can beneﬁt society, academic spin-offs

are often considered to be particularly valuable, since

successful startups can not only create new jobs and

increase tax revenues but also inspire other potential

founders to follow suit.

To promote knowledge transfer in general and the

https://orcid.org/0009-0000-1304-9012

https://orcid.org/0000-0003-4054-1306

https://orcid.org/0000-0001-7233-2547

creation of academic startups in particular, many uni-

versities employ innovation coaches that support re-

searchers by offering consulting services, by medi-

ating funding possibilities facilitated through indus-

try collaborations and investment partnerships, or by

arranging workshops and startup schools. However,

these activities are of a reactive nature, as they re-

quire researchers themselves, to seek out and seize

these offerings. To raise the awareness and to in-

crease the effectiveness of their knowledge transfer

activities, some universities have begun to take a more

proactive role. To do this, their innovation coaches

are regularly performing scouting activities to iden-

tify potential innovations and innovators inside their

organization. Conceptually, the innovation scouting

process can be split into three stages. In the ﬁrst stage,

the innovation coaches identify emerging trends by

gathering information about science and technology

in the early stages from both formal and informal

sources, including expert insights (Calvi et al., 2020).

Besides from talking to experts. the coaches usually

browse newsfeeds, projects, reports, scientiﬁc papers

as well as patents from inside and outside the orga-

nization. Once a trend is identiﬁed, the innovation

coaches ﬁnd the experts within the organization. To

do this, they usually rely on public information made

available through the websites of the different insti-

tutes, faculties and research groups. Finally, in the

last stage, the coaches prioritize the groups and indi-

Arzani, A., Handte, M. and Marrón, P.

Challenges in Implementing a University-Based Innovation Search Engine.

DOI: 10.5220/0012263100003598

In Proceedings of the 15th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2023) - Volume 1: KDIR, pages 477-486

ISBN: 978-989-758-671-2; ISSN: 2184-3228

477

viduals with matching expertise and results based on

their knowledge, motivation, founding potential and

collaboration skills (Albers et al., 2020) in order to

establish a contact with them.

Since there is no systematic approach to measure

the innovation potential of university members, the

scouting process is typically manual and relies heav-

ily on the experience of the innovation coaches as well

as their networking capabilities. Due to this, innova-

tion coaches are often networking experts and exhibit

an extensive network of industrial partners. However,

even if the innovation coaches can use their network

to easily identify enabling technologies and product

gaps, the success of scouting also depends on the un-

derstanding of organizational hierarchy and its under-

lying individuals. This makes the systematic identiﬁ-

cation, of the experts and the ranking and evaluation

of the innovation potential inside an organization a

difﬁcult challenge.

In this paper, we motivate the need for INSE

(INnovation Search Engine) to support innovation

coaches during their search for innovation potential.

While it may not be possible to completely system-

atize the innovation scouting process, we believe that

innovation coaches can greatly beneﬁt from an eas-

ily accessible and up-to-date contextualized view of

their organization. After discussing the information

needs of the scouting process, we outline a basic sys-

tem architecture to support it and we identify sev-

eral challenges. Our aim is to motivate vigorous

research in this area by illustrating the need for a

novel, data-driven approach to support the scouting

process which could further improve the effectiveness

of knowledge transfer.

The remainder of the paper is organized as fol-

lows: Section 2 introduces the related work in this

area; Section 3 outlines the information needs of the

innovation coaches during the scouting process. Sec-

tion 4 presents the INSE architecture in support of the

scouting process and Section 5 describes the identi-

ﬁed research challenges in implementing INSE for a

university. Finally, in Section 6 we conclude the paper

with a short summary and an outlook.

2 RELATED WORK

Academic entrepreneurship plays a pivotal role in

advancing scientiﬁc progress by bridging the gap

between research outcomes and real-world applica-

tions through spin-offs and has been the focus of

many studies in the research community (Rippa and

Secundo, 2019; Schultz, 2021; Kalinowski, 2016).

While (Rippa and Secundo, 2019) indicates and ad-

dresses the lack of in theoretical development in this

domain by underlining the necessity for more inclu-

sive studies that encompass technological, economic,

and social dimensions of academic entrepreneurship,

(Kalinowski, 2016) describes its weaknesses in exis-

tent research transfer systems by an analysis of scout-

ing implementation experiences in Polish universi-

ties. Thereby, technology scouting is described as

a strategic response to address the limitations of the

commercialization system by involving a systematic

approach to gather information in the realm of sci-

ence and technology and seeking bidirectional explo-

ration for novel opportunities in speciﬁc technological

ﬁelds. Furthermore, technology scouting and its qual-

ity are also emphasized by an analysis of academic

spin-offs of University of Potsdam (Schultz, 2021) as

a beneﬁcial factor that should be combined with other

systematic scouting activities in order to create a sus-

tainable raise in academic entrepreneurship.

Technology scouting is a strategic approach that

involves identifying and incorporating innovations,

such as those from startups, university intellectual

properties and acknowledges the increasing complex-

ity of products and services, recognizing that com-

panies and universities can not solely rely on inter-

nal efforts to innovate effectively (Wang and Quan,

2021). Therefore, many works include technol-

ogy/innovation scouting frameworks to promote star-

tups and innovation either in universities or compa-

nies through tools for technology/knowledge trans-

fer such as technology radars(Rohrbeck et al., 2006;

Golovatchev and Budde, 2010; Desruelle and Nepel-

ski, 2017; Berndt and Mietzner, 2021).

For instance (Rohrbeck et al., 2006) discusses

the use of technology scouting in Deutsche Telekom

Labs, where they employ a technology radar ap-

proach to enhance traditional newsletter-based scout-

ing. This method facilitates communication between

experts and scouts, improving the scouting process

through networking efforts. It involves an interna-

tional scouting network that analyzes trends, publica-

tions, and patents to identify potential technologies.

The scouts manually assess these technologies, and

an expert panel evaluates and ranks them, generat-

ing a report as the output. Additionally, as an ex-

tension (Golovatchev and Budde, 2010) introduces

innovation radars, which focus on market aspects of

technology and rely on reactive expert inquiries for

technology identiﬁcation and evaluation. Prior re-

search primarily emphasizes accelerating innovation

within corporate settings through technology and in-

novation radars. While their methodologies can also

be adapted for application within university contexts

to aid future startups, the authors of (Berndt and Mi-

KDIR 2023 - 15th International Conference on Knowledge Discovery and Information Retrieval

478

etzner, 2021) focus on digitizing these radars, con-

sidering them as tools to facilitate knowledge and

technology transfer in academic setting. They show-

case their outcomes through a web-based technology

radar, serving as an online collaborative tool. Their

argument centers on the absence of systematically ex-

plorative insight for technology scouts and highlights

how employing these tools signiﬁcantly simpliﬁes the

complexity of their evaluations.

There have been studies examining the role

of patents in advancing academic entrepreneurship.

These investigations have shown that the incorpora-

tion of knowledge transferred from the parent uni-

versity and academic founders through patents has a

noticeable impact on the success of Academic Spin-

Offs (Ferri et al., 2019). In fact some works put

their focus on utilizing patent analysis to support

the technology/innovation scouting process by iden-

tifying technology trends or new business opportu-

nities (An et al., 2018; Lee and Lee, 2017; Chen

et al., 2015). The work of (An et al., 2018; Lee

and Lee, 2017) uses text mining to derive keywords

on the technology hotspots from the patent content,

while (Chen et al., 2015) combine this technique with

piecewise linear representation of patent publications

throughout time to explore their hidden technology

trend patterns.

The aforementioned approaches prove the appli-

cability and importance of technology and innovation

scouting for the identiﬁcation of academic startups

through systematic frameworks and patents. How-

ever, these approaches lack the systematic data-driven

approach, as some rely heavily on manual human

interactions, or in the case of patent-oriented tech-

nology scouting only consider patents as the main

drivers of innovation. Innovation scouting for the cre-

ation of academic spinoffs should consider various

data sources aside from patents such as publications

or media along with data-driven processes involving

the identiﬁcation of university expertise. Despite the

theoretical depth of academic entrepreneurship in re-

lated work, there is clearly a gap for implementing

a data-driven solution for identifying the innovation

potential at universities.

3 INFORMATION NEEDS

As outlined in the introduction, innovation scouting

can be split into three stages, namely the identiﬁca-

tion of technology trends, the search for expertise and

the assessment of innovation potential. Each stage in-

volves the collection of speciﬁc data that is eventu-

ally used as the basis for decisions. Thus, scouting

not only encompasses data gathering but also requires

an interpretation that is speciﬁc to each stage. In the

following, we discuss the goals of each stage and de-

rive the resulting information needs. Although, our

description follows a top-down approach that allows

an innovation coach to identify candidates that match

a particular trend, it is noteworthy to point out that,

in practice, an innovation coach could also follow a

bottom-up approach that tries to determine how well

a certain researcher or research result matches a trend.

3.1 Identiﬁcation of Technology Trends

Knowledge about the current technology trends is an

important input for innovation scouting. Although,

some innovations may establish new trends by them-

selves, it is often easier to ﬁnd partners or acquire seed

funding, if the intended innovation domain matches or

extends an existing trend. Thus, the goal of this stage

is to monitor and understand the current and emerging

technology trends within different markets and indus-

tries. Thereby, the coach gains insights into what the

market demands, where opportunities lie, and which

directions innovation might take.

Identifying the technology trends involves re-

search and analysis of various factors such as con-

sumer preferences, cultural inﬂuences, and technical

advancements, to name a few. This means that inno-

vation coaches must follow the latest advancements in

the target sector. To do this, they can tap into regular

news sources (e.g., to cover consumer preferences or

cultural inﬂuences) as well as publications and patents

(e.g., to cover technological developments). In addi-

tion, they must relate the advancements with past de-

velopments to judge their signiﬁcance. Since this is a

complex undertaking that requires experience, inno-

vation coaches often have an extensive industry back-

ground and are in active exchange with their network

of collaboration partners.

3.2 Search for Expertise

Given the knowledge about the technical trends, in-

novation scouting must match the trends with the ex-

pertise inside the organization. For a university, this

could be a certain faculty, an institute, a research

group or a particular researcher that have theoreti-

cal and practical knowledge of the subject. To de-

termine the expertise of these entities, an innovation

coach must screen their research activities and results.

The latter can be extracted by analyzing the publica-

tions or patents created within the organization. For

the former, it is possible to look at grants awarded to

the different entities in the organization and the asso-

Challenges in Implementing a University-Based Innovation Search Engine

479

ciated research projects in which they are involved.

Besides from determining which entities exhibit

expertise in a particular domain, it may also be nec-

essary to judge whether the level and type of exper-

tise can serve as a basis for technology transfer. To

clarify this, consider that in a large organization such

as a typical university, the different entities are often

not focused on a single problem but are working on

a broad range of topics. Thus, although an entity is

working in a particular domain, it might not be its pri-

mary focus. Similarly, while some entities might be

addressing problems that exhibit an immediate rele-

vance to industry, other entities might be focusing on

addressing basic research problems that will only be-

come relevant for technology transfer in the future.

3.3 Assessment of Innovation Potential

After identifying a technology trend and ﬁnding the

relevant expertise within the organization, the goal of

the last step is to determine the set of persons that

should be contacted by the innovation coach. Thus,

given a set of entities exhibiting the targeted type and

level of expertise, it is necessary to identify the most

likely persons to participate in or drive the creation

of a startup. Towards this end, an innovation coach

must ﬁrst identify the relevant persons associated with

an entity and then assesses their background, skills,

knowledge, and abilities in order to prioritize them.

To identify the persons associated with an en-

tity, an innovation coach can analyze the organiza-

tion structure. To create a ranking, the coach can

then compare the background information of differ-

ent researchers within the organization. This might

include, for example, the CVs of the researchers, their

publication lists or their patent portfolios. In addi-

tion, the coach might want to compare the grants and

projects or look at the collaboration networks of the

researchers.

4 INNOVATION SEARCH ENGINE

As described previously, each stage of the scouting

process requires both data gathering and interpreta-

tion. For data gathering, innovation coaches must sift

through a broad range of information sources.

Examples include newsfeeds, research projects,

reports, grants, scientiﬁc literature, and patents, both

internal and external to the organization. Depending

on the stage, they must either organize the data over

time, e.g., to determine trends, or they must link the

information with the different entities within their or-

ganization, e.g., to determine the level and type of ex-

pertise of a particular entity inside the organization.

For interpretation, innovation coaches must aggregate

the data from different sources and analyze the re-

sults. This includes the comparison of aggregations,

e.g., to rank technology trends or to prioritize persons

that should be contacted. At the present time, there is

no thorough system-support for innovation scouting

at universities. As a result, data gathering is a compli-

cated and labor-intensive task. For information that

is directly available via the World Wide Web, inno-

vation coaches can use search engines such as Bing

and Google as their entry point. For restricted infor-

mation, such as (non-open access) publications or in-

ternal information about the organization, the coaches

must individually access a potentially large range of

information systems. While most of these systems

typically exhibit some way of searching for a partic-

ular piece of information, they usually lack the nec-

essary aggregation capabilities that are required for

innovation scouting. Thus, besides from searching,

the data gathering usually involves manual aggrega-

tion within and across different data sources.

A similar argument can be made about data inter-

pretation. Given that there is no systematic approach

to measure the innovation potential, the data interpre-

tation required to identify trends or to prioritize indi-

viduals usually relies heavily on the experience and

networking capabilities of the innovation coaches. As

a result, the decision-making during innovation scout-

ing is a rather subjective process. While it may not

be necessary to completely systematize the scouting

process, we believe that it can be greatly simpliﬁed

by providing adequate system-support.

To this end, we started the development of INSE

(INnovation Search Engine), an application-speciﬁc

search engine that aims to provide thorough support

for all three stages of innovation scouting. INSE

automatically gathers information from relevant data

sources such as newsfeeds, research projects, reports,

scientiﬁc literature, and patents, both internal and ex-

ternal to the organization. Then INSE stores and an-

alyzes the data. Thereby, it extracts relevant infor-

mation and links the different pieces of information

across different data sources. Using the linked in-

formation, INSE offers data aggregation, search, and

browsing functionalities to the innovation coaches.

On top of this, INSE offers a range of metrics that

support the three stages of innovation scouting. The

goal thereby is not to replace the (expert) opinions of

the innovation coaches with metrics, but to provide

them with a compact, up-to-date, and contextualized

view of the organization. Figure 1 shows a high-level

architectural overview of INSE. As part of an ongoing

research project, we are in the process of implement-

KDIR 2023 - 15th International Conference on Knowledge Discovery and Information Retrieval

480

ing the different components of INSE in order to sup-

port the innovation coaches of our university. In the

following, we specify each component and describe

their corresponding inputs and outputs.

Figure 1: Architecture of Innovation Search Engine.

4.1 Data Retrieval

The data retrieval component is responsible for the

gathering of data from various data sources. This

component aims to minimize the need for manual data

gathering by storing relevant data from multiple in-

formation systems in a single data repository. The

data retrieval process occurs through direct access to

the databases, APIs, as well as web crawling. For in-

stance, while a university maintains a library database

for the scientiﬁc publications of a certain researcher,

the information of their patents, participated research

grants, faculty, research group or institute might not

either be stored in a single database, or it is often scat-

tered among many databases. Therefore, the data re-

trieval merely depends on the type of data source.

The target data sources include (1) internal data

sources of the targeted university; (2) external data

sources such as scientiﬁc libraries for research pa-

pers, patent repositories, industry reports and news

portals. While the internal data sources reﬂect the

targeted university’s relevant information for the in-

novation coaches, the external data sources provide

contextual information as basis for comparison.

For internal data sources, INSE handles data gath-

ering via integrated connectors that have been devel-

oped to support different input formats. In addition,

INSE crawls the public website of the university to

gather information about the organizational structure

and the research activities. To do this, INSE starts

from a set of URLs given as an input parameter and

traverses the graph of pages by following the links

contained in pages. Since the pages may link to other

pages outside the university, the traversal can option-

ally be restricted by a list of target domains. During

the traversal of the graph, INSE automatically extracts

the web pages of the organizational hierarchy and its

underlying entities along with the attached metadata.

These entities may include information on institutes,

faculties, chairs, researchers research papers, funding

projects, news, and patents.

For external data sources, INSE gathers data ei-

ther via the APIs offered by the data provider or via

web crawling. Some of these sources that we have

added to INSE include libraries of patents such as

European patent ofﬁce (EPO), United States Patent

and Trademark Ofﬁce (USPTO), scientiﬁc portals

(Scopus, Web of Science, DBLP), industry reports

(Crunchbase) and news (Google Trends).

The actual data gathering strategy depends on the

source. For instance, in case of patent libraries, EPO

offers patent records that can be directly accessed

through an API, while the USPTO provides bulk

records that can be crawled. After gathering the data

from internal and external repositories, INSE stores,

indexes the data in a local repository for later search

and browsing.

4.2 Data Integration

The data integration component has two main goals,

namely entity extraction and linking of relevant in-

formation across different data sources. The gath-

ered data in the previous step includes structured and

unstructured information. While extracted informa-

tion from available data repositories (e.g., databases,

APIs) deliver organized ﬁelds and data formats, this

Challenges in Implementing a University-Based Innovation Search Engine

481

is not the case for the uncurated crawled data.

INSE employs Natural Language Processing

(NLP) pipelines to preprocess the crawled data.

Thereby, INSE performs data cleaning of stored un-

structured text data through the removal of language-

speciﬁc stop words and Unicode characters, as well

as lemmatization and stemming. This step cleans and

normalizes the contents of the local repository. Af-

ter cleaning and normalization, INSE aims to extract

the entities within the local repository. INSE utilizes

the Entity Recognition techniques (NER) to identify

and mine entities such as faculties, institutes, groups,

researchers, papers, patents, and news. Given that

the identiﬁed entities hold references to their content,

INSE utilizes their content to disambiguate them to

increase the quality of entities. For example. enti-

ties such as Thomas M

uller, M

uller Thomas, M

uller

Th. should point to the same entity depending on their

content, faculty, or research group.

After the identiﬁcation of the entities, each entity

needs to be further described by the corresponding

metadata. For instance, in the case of a researcher, in-

formation of their patents, publications and mentions

in news articles or industry reports should be linked

with and integrated into their proﬁle. INSE realizes

this through linking of entities across gathered inter-

nal and external sources that are stored in the local

repository. As its ﬁnal output, the integration com-

ponent produces a knowledge repository, holding the

extracted entities and their linkage to other entities.

4.3 Search and Browsing

The search and browsing component serves as an en-

try point to the innovation search engine. This compo-

nent offers active-search and explorative-search inter-

faces. It handles data aggregation, search, and brows-

ing functionalities for the innovation coaches.

By using the active-search interface, the innova-

tion coaches query and navigate through extracted en-

tities and technology trends in industry, research, and

news articles. Thereby, INSE can provide an in-depth

overview of expert proﬁles based on entity categories

or distinct technologies through dashboard visualiza-

tions for the user interface. Search queries offer in-

sight of the organizational information such as fac-

ulty or research group of an individual, as well as the

linked scientometrics and bibliometrics about the re-

search activities and their results. Moreover, the dash-

boards present information on the collaborations be-

tween researchers, research/industry project partners,

that may unveil knowledge clusters and potential col-

laborations. The active-search interface also taps into

the gathered data from external sources and demon-

strates the comparative analysis of technology trends

to both inside and outside the university.

The explorative-search interface offers the same

set of functionalities in an exploratory fashion. This is

helpful for innovation coaches who require a general

overview of the university entities and their activities

or may not search for a speciﬁc technology area or in-

dividual. The explorative dashboards show the active

technology trends to both inside and outside the uni-

versity and its researchers. Also, the explorative dash-

board also aggregates data and visualizes pointers to

the most active entities and their expertise in and outer

university, based on number of papers, patents, news

mentions.

Furthermore, INSE offers user interactions in the

search component to tag invalid or missing links in or-

der to give feedback to the data integration component

regarding the quality of entities and their linkage, so

that they can be further adjusted. Finally, the innova-

tion coaches can use the search interfaces, to generate

and bookmark a set of candidates, which serve as the

output of the search component. Depending on the

stage of the scouting process, these candidates consist

of selected technology trends, researcher(s), groups or

collaboration clusters.

4.4 Interpretation

Given a set of experts and technologies, the goal of

the interpretation component is to provide metrics on

the technology trends and individual(s) by measur-

ing the innovation potential of a particular technol-

ogy or individual(s). This is done in three steps that

match the stages of the scouting process described in

Section 3. (1) modelling technology lifecycles; (2)

prioritizing experts; (3) modelling innovation poten-

tial. To tackle the ﬁrst step, INSE aims to generate

the technology lifecycle to measure the technology’s

readiness and maturity. Thereby, INSE leverages the

stored local repository to analyze patterns of the his-

torical data on past emerging technologies and break-

throughs through patents and publications as well as

hype from media and news. As a result, INSE classi-

ﬁes the trendiness of a technology based on a mathe-

matical model that tries to capture the various phases

of the technology lifecycle.

Then in the second step, INSE prioritizes the ex-

perts based on the level of expertise for each of the

selected technologies, by taking into account the ex-

pert’s conference and journal impact score as well

as citation metrics for their publications and patents.

This provides a ranked list of experts that are active in

a particular domain.

As a last step, since being an expert in a spe-

KDIR 2023 - 15th International Conference on Knowledge Discovery and Information Retrieval

482

ciﬁc ﬁeld does not necessarily indicate founding po-

tential, INSE leverages the stored local repository to

analyze patterns of the historical data on academic

founders. Thereby, INSE utilizes a founder database

extracted from technology portals such as Crunchbase

and its linkage to the entities in the local repository to

compile a list of founders and non-founders (Arzani

et al., 2023). By generating a dataset of features

for founders and non-founders, patterns of founding

potential are modeled with a discriminative machine

learning algorithm such as Random Forests or a bi-

nary classiﬁcation deep learning model. Example for

a set of features could include number of patents,

publications or linked patents to industry or funded

projects as well as citation metrics. This allows INSE

to measure the innovation potential for each of the in-

dividuals in the prioritized list of experts.

The output of the interpretation component con-

sists of the lifecycles of the candidate technologies

and a list of prioritized experts, as well as the list of

high potential startup founders.

4.5 User Interface

The user interface builds up on the output of interpre-

tation component and consist of the well-timed tech-

nologies, ranked experts and high potential innova-

tors. This component also describes the main three

tasks of the innovation scouting process, that are re-

alized through the previous components. The UI gen-

erates reports for the identiﬁed technologies and the

selected candidates in the search component. Further-

more, the component offers interactive dashboards

and visualization to the innovation coaches via web

interfaces on the phases of technologies and the pro-

ﬁles of the selected candidates. The UI also visual-

izes bar charts on the ranked experts, their assessment

metrics as well as their estimated founding potential.

Additionally, the web interface explains a

roadmap of made decisions in the scouting activities

to the innovation coaches that led to the presented

results. This generates an overview of the selected

technology trends, researchers, and their collabora-

tion network in the search component. Also, the UI

shows the transition to the interpretation component

by including a description of applied models and their

parameters to offer explainability and future model

improvements based on the coach’s feedback. Ulti-

mately the innovation coach takes their own initiative

and decides whether they want to establish contact for

a startup opportunity.

5 CHALLENGES

During our ongoing work on implementing the INSE

architecture described above to support innovation

scouting at our university, we identiﬁed three main

challenges. The ﬁrst challenge is data acquisition

that describes the difﬁculties of data gathering in the

context of innovation scouting. The second challenge

refers to data integration that describes the issues re-

sulting from the need to link information from multi-

ple data repositories into a coherent knowledge repos-

itory. Lastly, the challenge of data interpretation un-

derscores the analysis and data modelling that aggre-

gates the available data into assessment metrics and

actionable intelligence. In the following, we provide

discussion on each of these challenges.

5.1 Data Acquisition

The scouting process requires access to multiple data

sources that include newsfeeds, research projects, re-

ports, grants, scientiﬁc literature, and patents, both in-

ternal and external to the university, whereas the ex-

ternal data sources determine a university’s expertise

and role (e.g., leader, follower) in a trending topic.

Internally, the challenge arises from the absence

of a central repository at the university housing all

the required data. As a result, the required data is

often spread across several systems that are operated

by different entities inside the university. Thereby, it

is noteworthy to mention that despite being present

in some system, some data might not be accessi-

ble or might even be deliberately blocked due to le-

gal or privacy reasons. Examples of systems that

are maintaining relevant data include ﬁnancial sys-

tems (e.g., for grant details or employee lists), library

systems (e.g., for publications) and web pages (e.g.,

for research activities). For the latter faculties or re-

search groups often maintain and update their own

content management systems that may not only ex-

hibit different structures but also present information

such as the research focus or ongoing research activ-

ities at different levels of detail. For example, some

groups might include information on the staff, publi-

cations, and projects but not necessarily bibliometrics

on publications or even patents or news. This means

that the missing information must be compensated

through internal sources (e.g., university-wide news

feeds) or external data sources (e.g., patent ofﬁces).

Thus, maximizing the data availability can become

quite challenging, since custom database connectors

and crawlers might have to be speciﬁcally adjusted to

each data source available inside the university.

The challenge of data acquisition is mirrored

Challenges in Implementing a University-Based Innovation Search Engine

483

externally, where relevant publication and patent

sources are dispersed among various publishers or

trademark ofﬁces. Data such as patents and publica-

tions are also available in scientiﬁc libraries behind

paid API calls that pose query limitations.

5.2 Data Integration

After acquiring the data from multiple sources, the

data must be linked to support the innovation coaches

by generating information about researchers’ proﬁles,

faculties, and groups. This involves the integration of

patents, publications, and projects with speciﬁc indi-

viduals and organizational units. This process neces-

sitates creating connections between the gathered data

sources and the internal data infrastructure of the uni-

versity. Before linking, these entities ﬁrst have to be

identiﬁed and disambiguated throughout all the avail-

able data sources. The disambiguation is a common

challenge in cases where there are researcher entities

with similar names or afﬁliations. This also includes

distinguishing between multiple researchers with the

same name or identifying variations in naming con-

ventions across different publications and databases.

To this end, general efforts i.e., use of an Open Re-

searcher Contributor Identiﬁcation (ORCID) or other

advanced techniques, such as name disambiguation

algorithms, network analysis, and afﬁliation disam-

biguation are necessary to ensure accurate and distinct

researcher proﬁles.

5.3 Data Interpretation

After retrieval and integration of multiple data sources

into a single repository and generating the linkage be-

tween the entities, INSE needs to interpret the pat-

terns of data via modelling in order to extract knowl-

edge and deliver actionable intelligence to the inno-

vation coaches. The challenges in data interpreta-

tion encompass three aspects for an effective scouting

process which we discuss in detail. The ﬁrst chal-

lenge refers to technology forecasting that signiﬁes

the study of technology trends through the analysis

of technology life cycles. The challenge of level of

expertise revolves around the ranking of the experts

based on their scientiﬁc proﬁle that helps the innova-

tion coaches to ﬁlter out the most suitable candidates.

Lastly, the challenge of founding potential under-

scores the evaluation of founding potential based on

academic background.

5.3.1 Technology Forecasting

Technology forecasting and technology trends

through a combination of news articles, patents and

scientiﬁc publications has been proven effective and

has taken the attention of many researchers (Chen

and Han, 2019; Asooja et al., 2016; Winnink et al.,

2019). Indeed, a clear picture of technology life

cycles is required to measure the impact of emerging

technologies that facilitates the innovation coaches’

ability to anticipate technological breakthroughs for

upcoming startups. The challenge here is to deﬁne

the maturity and readiness of a technology. Thereby,

a novel baseline for the maturity of technology

should be developed based on R&D activities such

as scientiﬁc literature, patents as well as social

media. Deﬁning this baseline provides a better

understanding of technology life cycles. One of

the ways to discuss the phases of a technology’s

is the Gartner hype cycles (Chen and Han, 2019).

Many scholars investigated the methodology of the

Gartner hype cycle in their work (Dedehayir and

Steinert, 2016; Van Lente et al., 2013). While the

authors of (Dedehayir and Steinert, 2016) provide

evidence that the Gartner model is a combination of

hype level through media with the business maturity

curve also known as S-curve, (Van Lente et al., 2013)

argue for the lack of mathematical foundation of the

Gartner model. Furthermore, Gartner hype cycle is

based on quantitative analysis, as there is no single

measure through surveys, evidence and forecasts,

meaning that the cycle involves expert judgment,

which can be quite bias. There are also reports of

discordance between the generated Gartner models

for some years and the actual trend of technologies

based on scientiﬁc literature and patents (Chen and

Han, 2019). Considering this challenge provides

an opportunity for further in depth analysis based

on bibliometrics and scientometrics to improve the

explainability of technology hype cycle models.

5.3.2 Level of Expertise

After determining the technology area and identify-

ing the knowledge experts or faculties at a university,

another factor of successful knowledge transfer is the

level of research done by the researchers. This in-

cludes the weight of parameters such as conference

and journal rankings, as well as impact factors that

are dependent on the type of disciplines.

The evaluation of research data through confer-

ence and journal rankings, along with impact fac-

tors, carries varying weight across different ﬁelds of

study. Depending on the discipline, the prestige of

known conferences and journals can directly corre-

late with technological advancements. Also, the im-

pact factor of a conference or journal also serves as

a potential metric for academic success based on the

main discipline. These metrics vary across domains,

KDIR 2023 - 15th International Conference on Knowledge Discovery and Information Retrieval

484

since in certain ﬁelds, innovative ideas and concepts

may emerge from less conventional channels. For

example, while journals in the medical ﬁeld often

have relatively high impact factors, the same can be

said about international conferences in the engineer-

ing discipline. Finding a valuable metric through

evaluation of venues and their impact factors provides

a list of high ranking experts inside a university.

5.3.3 Founding Potential

Universities increasingly encourage the involvement

of their academics in the transfer of knowledge to

the marketplace through spin-off activities (Siegel and

Wright, 2015; Gonz

alez-Pern

ıa et al., 2013). Study

based on a comparative analysis between Japanese

and US startup scene, indicates that the R&D activ-

ities serve as an important factor in entrepreneurship

activities (Kegel, 2016). Indeed, various researchers

(Farre-Mensa et al., 2016; Cagnani et al., 2022; Conti

et al., 2013; Helmers and Rogers, 2011) also argue

a direct correlation speciﬁcally between patent ac-

tivities as one of the main drivers of innovation and

founding startups. Therefore, considering the univer-

sities as a vast R&D network, a data-driven scouting

process can model the patterns of technology transfer

that come from patenting, scientiﬁc publications and

research grants.

Whereas previous studies for innovation identiﬁ-

cation, mainly depend on surveys and empirical re-

sults from individual institutions (Chung, 2023; Mon-

tebruno et al., 2020; Rivera-Kempis et al., 2021;

Sabahi and Parast, 2020), there is a gap that has to

take academic activities into consideration for inno-

vation scouting at universities. The challenge here is

to tap into the academic information of researchers to

identify existing founding potential. Thereby, there

is a need of list of previous founders which requires

access to specialized startup data sources. Also, the

startup data sources do not necessarily hold academic

information of their founders, therefore the founder

proﬁles have to be linked with other academic data

sources. By building a dataset of previous founders

and their academic behavior, the founding potential

of future founders can be measured and recommended

to the innovation coaches. However, there is no data

source that links the founder information to their aca-

demic background at the moment.

6 CONCLUSIONS

Technology transfer plays an important role in univer-

sities for the joint development and dissemination of

knowledge as a product that beneﬁts society through

innovation. In order to facilitate knowledge transfer,

many universities hire innovation coaches that em-

ploy a scouting process to identify faculty members

and students who possess the requisite knowledge,

expertise, and potential to establish startups. Since

there is no systematic approach to measure the inno-

vation potential of university members based on their

academic activities, the scouting process is typically

subjective and relies heavily on the experience of the

innovation coaches.

In this paper, we motivate the need for an Innova-

tion Search Engine as an integrated solution to sup-

port the scouting process. To do this, we described

the information needs of innovation coaches, and we

presented an architecture to cover them. Based on our

ongoing work on implementing the architecture, we

identiﬁed three challenges in order to motivate vigor-

ous research in this area. Our goal for future research

is to focus on data interpretation by building and eval-

uating models for technology hype cycles, as well as

ranking and innovation potential of researchers.

ACKNOWLEDGEMENTS

This work has been funded by GUIDE REGIO which

aims to improve the ability of the science support cen-

ter of the University of Duisburg-Essen in the iden-

tiﬁcation, qualiﬁcation, and incubation of innovation

potentials.

REFERENCES

Albers, A., Hahn, C., Niever, M., Heimicke, J., Marthaler,

F., Spadinger, M., et al. (2020). Forcing creativity

in agile innovation processes through asd-innovation

coaching. In Proceedings of the Sixth International

Conference on Design Creativity (ICDC 2020), pages

231–238.

An, J., Kim, K., Mortara, L., and Lee, S. (2018). Deriv-

ing technology intelligence from patents: Preposition-

based semantic analysis. Journal of Informetrics,

12(1):217–236.

Arzani, A., Handte, M., Zella, M., and Marr

on, P. J. (2023).

Discovering potential founders based on academic

background.

Asooja, K., Bordea, G., Vulcu, G., and Buitelaar, P. (2016).

Forecasting emerging trends from scientiﬁc literature.

In Proceedings of the Tenth International Conference

on Language Resources and Evaluation (LREC’16),

pages 417–420.

Berndt, M. and Mietzner, D. (2021). Facilitating knowl-

edge and technology transfer via a technology radar

as an open and collaborative tool. New Perspectives in

Challenges in Implementing a University-Based Innovation Search Engine

485

Technology Transfer: Theories, Concepts, and Prac-

tices in an Age of Complexity, pages 207–230.

Bliznets, I. A., Kartskhiya, A. A., and Smirnov, M. G.

(2018). Technology transfer in digital era: legal en-

vironment. Journal of History Culture and Art Re-

search, 7(1):354–363.

Cagnani, G. R., da Costa Oliveira, T., Mattioli, I. A.,

Sedenho, G. C., Castro, K. P., and Crespilho, F. N.

(2022). From research to market: correlation between

publications, patent ﬁlings, and investments in devel-

opment and production of technological innovations

in biosensors. Analytical and Bioanalytical Chem-

istry, pages 1–9.

Calvi, R., Pihlajamaa, M., and Servajean-Hilst, R. (2020).

Innovation scouting: a new challenge for the purchas-

ing function. The nature of purchasing: Insights from

research and practice, pages 295–313.

Chen, H., Zhang, G., Zhu, D., and Lu, J. (2015). A patent

time series processing component for technology in-

telligence by trend identiﬁcation functionality. Neural

Computing and applications, 26:345–353.

Chen, X. and Han, T. (2019). Disruptive technology fore-

casting based on gartner hype cycle. In 2019 IEEE

technology & engineering management conference

(TEMSCON), pages 1–6. IEEE.

Chung, D. (2023). Machine learning for predictive model in

entrepreneurship research: predicting entrepreneurial

action. Small Enterprise Research, pages 1–18.

Conti, A., Thursby, J., and Thursby, M. (2013). Patents as

signals for startup ﬁnancing. The Journal of Industrial

Economics, 61(3):592–622.

Dedehayir, O. and Steinert, M. (2016). The hype cycle

model: A review and future directions. Technologi-

cal Forecasting and Social Change, 108:28–41.

Desruelle, P. and Nepelski, D. (2017). The’innovation

radar’: A new policy tool to support innovation man-

agement. Available at SSRN 2944104.

Farre-Mensa, J., Hegde, D., and Ljungqvist, A. (2016). The

bright side of patents. Technical report, National Bu-

reau of Economic Research.

Ferri, S., Fiorentino, R., Parmentola, A., and Sapio, A.

(2019). Patenting or not? the dilemma of academic

spin-off founders. Business Process Management

Journal, 25(1):84–103.

Golovatchev, J. and Budde, O. (2010). Technology and in-

novation radar - effective instruments for the devel-

opment of a sustainable innovation strategy. In 2010

IEEE International Conference on Management of In-

novation and Technology, pages 760–764.

Gonz

alez-Pern

ıa, J. L., Kuechle, G., and Pe

na-Legazkue,

I. (2013). An assessment of the determinants of uni-

versity technology transfer. Economic Development

Quarterly, 27(1):6–17.

Helmers, C. and Rogers, M. (2011). Does patenting help

high-tech start-ups? Research Policy, 40(7):1016–

1027.

Kalinowski, B. (2016). Increasing the potential for commer-

cialisation of innovation and research results within

polish universities. Modern Management Review,

21(23):2.

Kegel, P. (2016). A comparison of startup entrepreneurial

activity between the united states and japan. Journal

of Management Policy & Practice, 17(1).

Lee, M. and Lee, S. (2017). Identifying new business oppor-

tunities from competitor intelligence: An integrated

use of patent and trademark databases. Technological

Forecasting and Social Change, 119:170–183.

Montebruno, P., Bennett, R. J., Smith, H., and Van Lieshout,

C. (2020). Machine learning classiﬁcation of en-

trepreneurs in british historical census data. Informa-

tion Processing & Management, 57(3):102210.

Rippa, P. and Secundo, G. (2019). Digital academic en-

trepreneurship: The potential of digital technologies

on academic entrepreneurship. Technological Fore-

casting and Social Change, 146:900–911.

Rivera-Kempis, C., Valera, L., and Sastre-Castillo, M. A.

(2021). Entrepreneurial competence: Using machine

learning to classify entrepreneurs. Sustainability,

13(15):8252.

Rohrbeck, R., Heuer, J., and Arnold, H. (2006). The tech-

nology radar - an instrument of technology intelli-

gence and innovation strategy. In 2006 IEEE Interna-

tional Conference on Management of Innovation and

Technology, volume 2, pages 978–983.

Sabahi, S. and Parast, M. M. (2020). The impact of en-

trepreneurship orientation on project performance: A

machine learning approach. International Journal of

Production Economics, 226:107621.

Schultz, C. (2021). Does technology scouting impact spin-

out generation? an action research study in the context

of an entrepreneurial university. New Perspectives in

Technology Transfer: Theories, Concepts, and Prac-

tices in an Age of Complexity, pages 107–128.

Siegel, D. S. and Wright, M. (2015). Academic en-

trepreneurship: time for a rethink? British journal

of management, 26(4):582–595.

Van Lente, H., Spitters, C., and Peine, A. (2013). Com-

paring technological hype cycles: Towards a the-

ory. Technological Forecasting and Social Change,

80(8):1615–1628.

Wang, C.-H. and Quan, X. I. (2021). The role of ex-

ternal technology scouting in inbound open innova-

tion generation: Evidence from high-technology in-

dustries. IEEE Transactions on Engineering Manage-

ment, 68(6):1558–1569.

Winnink, J., Tijssen, R. J., and Van Raan, A. (2019).

Searching for new breakthroughs in science: How ef-

fective are computerised detection algorithms? Tech-

nological Forecasting and Social Change, 146:673–

686.

KDIR 2023 - 15th International Conference on Knowledge Discovery and Information Retrieval

486