STRATEGIC INNOVATION MANAGEMENT ON THE BASIS
OF SEARCHING AND MINING PRESS RELEASES
Jan Finzen, Maximilien Kintz, Holger Kett
Fraunhofer IAO, Nobelstr. 12, 70195 Stuttgart, Germany
Steffen Koch
Visualization and Interactive Systems Group (VIS), Universität Stuttgart, Universitätsstraße 38, 70569 Stuttgart, Germany
Keywords: Web Information Retrieval, Press Releases, Strategic Innovation Management.
Abstract: Press releases may contain a lot of information that is especially applicable in strategic innovation
management: They contain up-to-date information by definition and thus may give hints to upcoming trends
and techniques. They also tell a lot about the strategies of partners, customers, and, most of all, competitors.
We analysed many of today’s existing press release search engines and identified a number of
shortcomings: The query frontends do not provide enough flexibility with regards to search space
restriction, the result lists presentation typically cannot be influenced by the user, and the ranking order
stays often unclear. Press releases offer a number of features that make them useful for automatic handling
but are widely ignored by today’s search engines: They are relatively homogenously structured and contain
certain kinds of easy-extractable meta-data that can be utilized for use cases such as monitoring trends (date
of publishing), discovering geographical competency clusters (author and address information), or
identifying relations between companies (firm name mentioning). We describe the prototype of a new press
release search engine that makes use of the above-mentioned meta-data and additionally offers sophisticated
search features tailored to the needs of innovation professionals.
1
1 INTRODUCTION
Strategic innovation management deals with “the
planning, organisation, execution, and control of any
innovation activity aiming at the creation and
preservation of competitive advantages” (Goos and
Hagenhoff, 2003). A lot of an innovation manager’s
daily work is somehow related to searching the web
for information. A recent survey among industrial
innovation actors and experts states that web search
engines clearly dominate the internet-based
technologies used by innovation actors, followed by
online editions of professional journals (cf. Novanet,
2006).
Some types of documents, accessible online, are
of special value for strategic innovation management
because they are very likely to include innovation-
relevant information.
Press releases represent such a document type.
They are highly relevant within strategic innovation
management, as they can be expected to contain
some information that is “new” in one way or
another (in fact the term “news release” is often used
as a synonym). Press releases are usually authored
by organisations to spread some information
considered worth publishing and constitute a means
for public relations. They target media editors who
are expected to pick up the topic and develop articles
on it.
Press releases can tell a lot about activities and
plans of customers, partners, as well as competitors
and may thus be applied as a source of information
for business or competitive intelligence which we
see as sub area of strategic innovation management.
As of today, most press release portals offer
rather limited search functions and largely ignore the
intrinsic features of press releases such as location
and date information.
Within this paper we introduce a prototype of a
sophisticated press release search engine that takes
into account a number of innovation professionals’
requirements. The remainder of this paper is
structured as follows:
347
Finzen J., Kintz M., Kett H. and Koch S.
STRATEGIC INNOVATION MANAGEMENT ON THE BASIS OF SEARCHING AND MINING PRESS RELEASES.
DOI: 10.5220/0001840803470353
In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page
ISBN: 978-989-8111-81-4
Copyright
c
2009 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
In Section 2, we analyse the state of the art
regarding press release search engines. We identify
several shortcomings and propose methods to
overcome them within a meta search engine
approach based on RSS data feed integration.
In Section 3, we introduce the current state of
our search engine’s prototype and explain the most
important features using real world examples.
In Section 4 we discuss the results. We point out,
what problems still have to be overcome and give an
outlook to future activities.
2 METHODS AND MATERIALS
In this section, we first define the specific
information seeking ambitions of “innovation
officers”. Keeping this in mind, we analyse the state
of the art regarding press release search engines. We
identify several shortcomings and propose methods
to overcome them by means of a meta search engine
approach based on RSS data feed integration.
2.1 Target Group Analysis
There is no official definition of an “innovation
professional”. Within the context of our work, we
define the strategic innovation professional as a
specialist who is taking care of market development
and oversees the actions of customers, partners, and
competing companies. Regarding the innovation
officer’s information retrieval abilities, we account
him to be rather motivated in investing considerable
effort to retrieve high-quality results from search
tasks. Stock and Lewandowski (2006) distinguish
three different classes of search engine users:
1. Information professionals: Experts for
searching and finding documents with
professional background in information
sciences. They know relevant databases,
develop search strategies and use
professionally elaborated queries. Research is
predominantly done by order of a third party.
2. Professional end users: Business experts.
Searching because of their business-related
information needs. They know the suitable
business-specific data bases and search
engines, but rarely develop research strategies
and professionally elaborated queries.
3. Amateurish users: Searching mostly for
private and sometimes for professional reasons.
Restricting their efforts to query the surface
web using a general-purpose search engine
(e.g. Google). They do not apply systematic
search strategy and normally use short queries.
Innovation professionals do not match any of
those categories exactly but rather combine
attributes of the first two classes: They are experts
concerning their companies’ business. But they are
expected to provide a wider view and to be able to
“think outside the box”. Therefore, they bring along
a broad knowledge of where and how to find
information meeting their requirements and needs.
Innovation professionals perform searches in a
frequent and strategic manner – thus, they know how
to effectively use search engines.
2.2 State of the Art Analysis
Today, press releases are most often disseminated
using commercial press release distribution services.
Based on web research and experts’ opinions we
identified the most important German press release
portals and some of the largest US portals relevant
for our examination. We analysed these portals with
regard to the following aspects:
Business model: How do they earn money?
Amount of content: How many press releases
are published?
Search frontend features: How easily and
detailed can a query be specified?
Search result display: How are search results
presented to the user?
2.2.1 Business Model
Usually the business models of these services are
based on advertisement (almost all the services we
know use Google AdSense). In some cases, the
author of a press release has to pay for the
publishing. In these cases, the charges amount to an
average of 100 to 300 Euros per release, depending
on the diffusion rate and additional services like
click-tracking etc.
2.2.2 Amount of Content
The amount of content, i.e., the publishing activity
of the surveyed press release distribution portals
varies significantly. While only a few portals publish
up to 800 press releases on average each day, most
portals restrict themselves to less than 50 releases
each day. Thus, the distribution curve forms a
typical long-tail (cf. figure 1).
WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies
348
Table 1: Evaluation results of examined press release websites (excerpt).
Explanation of columns: Publishing costs per press release; Query frontend quality: ●●● = many options and
easy to use, = few options, hard to use; Result presentation quality: ●●● = very clear and complete
presentation, = few information, unclear presentation; RSS feed quality: ●●● = feed complete, = important
data missing, - = no feed available; Total query quality: mean of query frontend, result presentation, and RSS
feed quality; Amount of content: ●●● = 200 and above, ●● 100 to 200, = 50 to 100, = 0 to 50.
0
100
200
300
400
500
600
700
800
900
PR New
swire
Pr
esseEcho
PR
Web
Open
PR
PressePorta
l
IN
AR
Li
fePR
IDW
PM
-Webservice
Firmenp
ress
e
Catego
rynet
Online A
rtikel
DailyNet
Pres
sbot
India PR
Wir
e
AlphaG
alileo
Re
pandre.com
Press
eT
ext
Op
enBroa
dcast
Generalite.com
P
res
sText
Press1
C
om
muniqués de presse
USA
News
Figure 1: Long-tail distribution of amount of content.
2.2.3 Search Frontend Features
All the web sites we looked at offered basic search
functionality that allows searching for key words in
the full text of all press releases. Many sites
additionally offer an “expert search” that allows
further restriction of the search space by specifying
one or more of the following constraints:
The publishing time frame
Boolean combinations of key words,
Geographical regions (e.g., countries or states)
Company names, or
Categories (whereby the granularity of
classification systems varies significantly).
Remarkably, none of the web sites offer all of
those features. However, we consider all of these
constraints useful, especially if the user repeatedly
searches with professional background, as it is the
case within the strategic innovation management.
2.2.4 Search Result Presentation
While all portals present their search results as lists
which link to the complete text, as known by
standard search engines, the information displayed
in the list varies significantly, too: For example,
PRNewswire restricts itself to the title, the author’s
name and the date of publishing. LifePR also
displays the company name, category information,
the first sentences of the text and, if available, also
images. The number of results per page ranges from
10 to all. But in most cases this cannot be configured
by the user. The ranking algorithm often remains
unclear and can only rarely be influenced by the
user.
2.3 Examination Results
Table 1 shows an excerpt of our examination results
ordered by the amount of content. As many different
press release portals exist and press releases are
often exclusively published by one of them, a meta
search approach seems promising to increase the
Name Language
Publishing
costs (€)
Query
frontend
quality
Result
presentation
quality
RSS feed
quality
Total
query
quality
Press
releases
per day
Amount of
content
PR Newswire EN 120 ●○ ●○ ●●● ●● 800 ●●●
PresseEcho DE - 500 ●●●
PR Web EN 80-230 500 ●●●
OpenPR DE - 280 ●●●
PressePortal DE 315 ●● ●●● ●● 200 ●●●
INAR DE - ●● 175 ●●
LifePR DE 99 ●● ●● ●● ●● 130 ●●
PM-Webservice DE - ●○ 80
Firmenpresse DE - ●● 70
Online Artikel DE -
●○
50
DailyNet DE - 50
Pressbot DE - 50
AlphaGalileo EN-DE 50-100 ●● ●● ●● 25
PresseText DE 167 ●● ●○ 20
OpenBroadcast DE - 20
PressText DE 160 ●● ●● - 10
Press1 DE - ●● 5
STRATEGIC INNOVATION MANAGEMENT ON THE BASIS OF SEARCHING AND MINING PRESS RELEASES
349
recall and improve the ease of use. Both the query
frontend and the presentation of search results offer
room for improvement. Additional usage of meta
data for information aggregation, for trend
monitoring, or identification of competency clusters
is not provided at all in today’s press release portals.
However, these approaches seem quite promising for
innovation management, especially when they are
applied in combination with a meta search engine
approach.
3 AN IMPROVED PRESS
RELEASE META SEARCH
ENGINE
Based on our findings we designed and implemented
a press release meta search engine which,
is based on RSS feed aggregation,
follows a best-of-breed approach with regards
to the query and result frontend, and
offers additional search features based on meta
data like dates and addresses.
In the following we will give an overview of the
prototype’s current state and illustrate our search and
mining approaches using real-world examples.
3.1 Query Frontend
As we stated in the previous section, the query
frontends of today’s press release search engines do
lack the options for restricting the search space
according to the user’s needs. Thus, we defined an
expert query frontend that allows an arbitrary
combination of all the search space constraints listed
in Section 2.2.3 (see Figure 2).
Figure 2: Configuring the search space.
This expert query frontend is intended to be used by
motivated experienced users. In addition, simple ad
hoc queries can be entered using a query dialog with
reduced complexity.
3.2 Search Result Presentation
We followed a “best-of-breed”-approach when
designing the result list presentation. Unlike the
press release search engines we analysed, our new
prototype lets the user influence both the number of
results displayed per page and the ranking of the list
(see Figure 3).
Figure 3: Search results presentation.
As we follow a meta search engine approach the
found results may well come from different press
release distribution sites thus looking quite
inhomogenously when displayed on the original
sites. To improve readability we added a uniform
result display that shows each result in a similar look
and feel (see Figure 4).
Figure 4: Press release presentation.
WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies
350
Since innovation officers carry out long-ranging
search tasks quite regularly, search queries need to
be stored and automatically repeatedly carried out
within planned intervals. In our search engine, the
user may subscribe to any query she likes and obtain
the search results as RSS data feeds, again.
3.3 Meta Data Usage
Press releases, apart from the actual content, contain
additional information that can be utilised within the
search process.
Title: Can be used for duplicate detection,
which is quite important within a meta search
engine approach
Abstract: If it exists, it can be displayed within
the result list to improve usability and
supersedes algorithmic summarization
approaches.
Date: Allows restricting the search space and
time-tracking of press releases.
Author and address information: Apart from
restricting the search space, this information
can be utilised to group releases, e.g. by
company. Address information can be geo-
coded and allow sophisticated geographical
views.
Tags and categories: Allow grouping of search
results.
In the following we will introduce how our
prototype, which exploits this information to offer
advanced search features, is tailored to the needs of
strategic innovation management.
3.3.1 Monitoring Trends and Detecting
Events
The date of publishing is included in every press
release. Thus, it is quite easy to keep track of the
number of results for an arbitrary search query and
visualize the results as diagrams.
In Figure 5 we illustrate this concept by
contrasting the number of press releases of two
different companies. The red line indicates the
number of press releases published by apple between
July 1
st
and July 31
st
. Note the peek on July 10
th
which can be traced back to the fact that the iPhone
G3 was released on this very day. Similar peeks
were observed when the new iPod generation was
introduced in September 2008.
If combined with a keyword-based search for an
arbitrary topic, trends can be monitored. Figure 6
shows the temporal distribution for the keyword
“Finanzkrise” (financial crisis) between July and
November 2008.
Figure 5: Time-tracking press releases of companies.
Figure 6: Time-tracking the topic “financial crisis”.
We complemented the line chart view with a bar
chart view that visualises shorter time-frames better.
If the user clicks on the diagram the search results
for the underlying query restricted to the
corresponding time-frame are displayed as a list (see
Figure 3).
3.3.2 Identifying Geographical Competence
Clusters
Similar to the application of dates for visualising the
temporal distribution of press releases we can use
geographical information included in press releases
to depict geographical distributions. This feature
allows straight-forward identification of competence
clusters by showing them on maps. Figure 7 shows a
simple example: If we search for “software”-related
press releases we can easily identify a high
publication density south of the San Francisco bay
area which of course is due to many software
companies being located around the Silicon Valley.
For our example we combined the Google Maps
mashup service with a Quality Threshold Clustering
(Heyer et al, 1999) – based algorithm to detect an
arbitrary number of clusters.
STRATEGIC INNOVATION MANAGEMENT ON THE BASIS OF SEARCHING AND MINING PRESS RELEASES
351
Figure 7: Cluster Visualisation.
Another way of visualising geographical
distributions is colouring maps. This kind of
visualisation is often used to highlight differences in
distribution between geographical regions, e.g.
election results. We implemented this view on the
level of the German federal counties as well as of
the US federal states. Figure 8 shows the distribution
of press releases for the search term “umwelt”
(environment) within the counties of Germany.
Figure 8: Colouring Maps.
3.3.3 Identifying Relations between
Companies
Furthermore, we implemented an algorithm to
identify relations between two companies. For each
company that publishes a press release we find out:
Which other companies have mentioned this
company in their press releases and how
often?
Which other companies have been mentioned in
press releases of this company and how often?
Within the current implementation status we
only detect names of companies that have already
published at least one press release themselves – and
thus can be found in our data base. In future
releases, a well-maintained list of company names
will improve the approach.
Both - forward and backward citations - can be
visualised as reference graphs. Figure 9 shows a
reference graph for the company “Business
Objects”. Citation frequency is encoded by the
thickness of the corresponding edge. In figure 9 this
can be observed by the fact that the edge between
“Business Objects” and SAP is thicker, which
correlates to the fact that Business Objects is a SAP
company since the takeover in October 2007.
Figure 9: Company Reference Graph, Level 1.
The reference graph can be computed
recursively – but time constraints have to be taken
into account. The computation time highly depends
on the number of different companies stored in the
database and the number of referencing companies
identified in the previous step. Within the current
prototype we therefore restricted the reference graph
computation at runtime to the second level
(references of references). Nevertheless, the process
can be accelerated using offline precomputation
steps. As future improvement we will add heuristics
for firm name unification (cf. Magnani and Montesi,
2007), in order to detect that for example, “Apple”
and “Apple Comp. Inc” denote the same entity.
3.4 Implementation Notes
As nearly all of the analysed press release portals
offer RSS based data streams, feed aggregation
seems to be a promising means of integrating the
different data sources within a meta search engine
approach. On this basis, we started building a press
release corpus by aggregating the newsfeeds of the
five most active portals. Unfortunately, the provided
RSS data streams do not include the full text but
only a URL pointing to the HTML page. In some
cases, meta information is also not explicitly
marked-up. Thus, we built some additional page
WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies
352
scrapers to extract both, the full text and the meta
data from the website.
The current prototype consists of two
applications:
A standard Java application fetches the newest
press releases every few minutes by reading
the RSS feeds and web scraping additional
data for any newly retrieved message.
The search engine itself is implemented as a
standard Java web application running on an
Apache Tomcat web server.
All data is currently stored in a MySQL
database, but we will soon migrate to a Lucene
search index to improve retrieval performance.
4 DISCUSSION AND
CONCLUSIONS
Within this paper we introduced a prototype of a
new press release search engine that offers advanced
search possibilities by taking into account the
specific structure and meta data of press releases.
The findings so far indicate that systematic press
release observation can help exposing facts and
developments, which are highly useful within
strategic innovation management. Nevertheless the
examples we provided are still rather simple. To get
more interesting results, more and more carefully
selected data is needed.
One of our findings is that the publishing activity
among the available press release distribution portals
forms a typical long-tail. A way to improve the
search data corpus size thus lies in integrating as
many data sources as possible, i.e., not only the four
or five biggest portals have to be accounted for, but
also the many small providers.
The extraction and processing of meta data must
be enhanced; for instance, company names like IBM
and IBM Corp. are currently treated as two different
organisations. In future releases we intend to apply
named entity recognition algorithms to reliably
identify and unify company names. Additionally, the
usage of a well-maintained toponym reference list
will enhance the geo locating functions. The
classification of press releases currently depends on
explicit markup, which is not offered by all
providers. We are therefore evaluating a stochastic
topic detection approach to automatically classify
press releases based on their textual content.
To further improve the search result quality we
are currently adding a semantic-based query
expansion mechanism. This will increase the recall
of the meta search engine and improve the outcomes
of the trend monitoring and cluster identification
approaches - as both rely heavily on the amount of
relevant data retrieved.
The feedback we received from cooperating
companies so far support our assumption that the
tool is regarded useful by innovation professionals.
But more systematic user studies have to be
arranged to evaluate formally how useful the
implemented features are with regard to different
user types and different business areas.
As part of our further activities we will integrate
the press release search engine as one module
among others into a so-called innovation mining
cockpit. This cockpit will form a single point of
entry for all of the innovation professional’s web
search related activities (see Stathel et al, 2008, for
details).
REFERENCES
Goss, P., and Hagenhoff, S., 2003.
Strategisches Innovationsmanagement: Eine
Bestandsaufnahme. In Schumann, M. (ed.),
Arbeitsbericht Nr. 11/2003 des Instituts für
Wirtschaftsinformatik der Georg-August-Universität
Göttingen, Göttingen.
Heyer, L. J., Kruglyak, S., and Yooseph, S., 1999.
Exploring Expression Data: Identification und
Analysis of Coexpressed Genes. In Genome Res. 1999
9: 1106-1115.
Magnani, M., and Montesi, D., 2007. A study on company
name matching for database integration. Technical
Report UBLCS-07-15. May 2007.
Novanet, 2006. Information in der Internetökonomie. 2nd
newsletter of the NovaNet project, http://www.nova-
net.de/fhg/Images/nova-net_2-Newsletter_tcm231-
60869.pdf. Accessed September 9th, .2008.
Stathel, S., Finzen, J., Riedl, C., and May, N., 2008.
Service Innovation in Business Value Networks.
In Proceedings of the XVIII International RESER
Conference.
Stock, W. G., and Lewandowski, D., 2006.
Suchmaschinen und wie sie genutzt werden. WISU
35(2006)8-9, 1078-1083.
1
The project was funded by means of the German Federal
Ministry of Economy and Technology under the promotional
reference “01MQ07012”. The authors take the responsibility
for the contents.
STRATEGIC INNOVATION MANAGEMENT ON THE BASIS OF SEARCHING AND MINING PRESS RELEASES
353