=Paper=
{{Paper
|id=Vol-2620/paper8
|storemode=property
|title=A Linguistic Analysis of Startups in The Context of the Air Transport Industry Management
|pdfUrl=https://ceur-ws.org/Vol-2620/paper8.pdf
|volume=Vol-2620
|authors=Olga Zervina
|dblpUrl=https://dblp.org/rec/conf/balt/Zervina20
}}
==A Linguistic Analysis of Startups in The Context of the Air Transport Industry Management==
https://ceur-ws.org/Vol-2620/paper8.pdf
A Linguistic Analysis of Startups in The Context of the
Air Transport Industry Management
Olga Zervina1[0000-1111-2222-3333]
1
Transport and Telecommunication Institute, 1 Lomonosova
street, LV-1019, Riga, Latvia
zervina.o@tsi.lv
Abstract. Much research has studied how a company can maximize its profit.
Relatively small number of them focused on Value Proposition, though the
number of authors proved that companies that deliver multiple values experi-
ence better business performance. This article describes a current research on
linguistic analysis of startups in the context of the air transport industry. Ana-
lyzing startups manually is a very time consuming task, so the automation of
the process would be beneficial. The author takes corpus linguistic approach,
created an experiment protocol and is on the stage of conducting an experiment.
Under this experiment air transportation startups’ landing pages were collected
in the number of 800. 100 annotators first were preliminary surveyed and then
trained to annotate startups. Post-annotation training will be conducted to un-
derstand the difference in expertise level. The annotation results will be further
analyzed and linguistic features and patterns will be identified. As a result of
the research, an author will develop a methodology for analysis of values based
on a model of automatic identification of values in the text of a startup’s land-
ing page in the air transportation industry.
Keywords: Startups, Value Proposition, Air Transport, Annotation.
1 Introduction
Value chain analysis is a primary corporate strategy tool. Traditionally relatively few
values have been pursued; price, quality, etc. Recently, customers are demanding a
wide variety of values. Most importantly, some of the values that have previously
received little attention become dominant values (concept known as value shift).
Some examples: cars - eco-friendliness, electronics – usability, food - fairness, organ-
ic. As new values become prominent, industries and their companies will be trans-
formed; not following values could result in wasted resources (pursuing the wrong
values) and becoming irrelevant (not pursuing the right ones). Hence, the ability to
identify values early on is important for business performance. Startups are often the
first to discover new values. As an example, Skytran, startup from air transportation
field, offers autonomous, zero-emission vehicles arrowing above congested streets.
Copyright © 2020 for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0)
57
The company’s landing page identifies its values as “high speed, high capacity, low
cost” [1]. Analyzing startups manually is a very time consuming task. The author
assumes that one startup landing page takes 10 minutes to examine. Air transport
industry offers around 10000 startups per year, requiring 1 full-time analysts to keep
track of startups of just one industry on annual basis; and there are 100’s of intersect-
ing industries in aviation (food & beverage, hospitality, etc.). Methods that allow at
least in part automate this process would be beneficial. This research describes the
first stages of developing a methodology of analyzing values within industry based on
startups value propositions.
The main research question is tools and procedures for analysis of value proposi-
tion of startups in the area of air transportation from their textual description in the
frame of developing a methodology for analysis of values based on a model of auto-
matic identification of values in the text of a startup’s landing page in the air transpor-
tation industry.
The initial contribution of this research is the development of methodology of
analyzing values within industry based on startups value propositions. The procedures
and tools integrated in the model are either articulated or developed by the author
throughout the Thesis.
The approach. The author uses a Natural Language Processing (NLP) approach to
identify the methods and features well suited for this problem. A bottom-up (data-
driven) technique is taken, i.e. the author first constructs the dataset and then analyzes
it using both computational and linguistic methods to identify which features and
methods perform the best. The author also takes the corpus linguistic approach, i.e.
reliable language analysis is more feasible with corpora collected in the field in its
natural context.
The object of the research is startups’ landing pages in the field of air transport in-
dustry.
The subject of the research is value proposition of a startup.
The tasks to be performed are the following:
1. To study theoretical literature on the 5. To identify features (e.g. linguistic,
research topic semantic)
2. To examine existing research on the 6. To identify patterns
research topic 7. To develop testable predictions
3. To collect preliminary data 8. To build a model
4. To conduct an experiment 9. To test a model
2 The Methodology and Methods of Research
The research methodology is presented in Figure 1. The author takes data-driven
scientific method. The general idea was adopted from the paper describing the scope
of big data in medicine [2].
58
Fig. 1. Stages: (A) Framing the problem and general hypotheses. (B). Data collec-
tion and exploratory experimentation/analysis. (C) Formulation of specific hypothe-
ses. (D) Testing the hypotheses. (E) Accepting or rejecting the hypotheses.
The research hypotheses were formulated as follows:
general hypothesis: values could be automatically identified from text (landing
page of a startup)
specific hypotheses: 1.Significant increase in number of known values is possible
even for people with higher level of domain familiarity; 2. Linguistic features can
improve accuracy of value identification.
59
With the accessibility of large datasets and advanced correlation/statistical tools,
will we still need to rely on hypotheses in scientific research? Traditionalists say that
in purely data-driven methods, one may not realize where to look for those riveting
findings if no hypotheses were formed beforehand.
Big data advocates, on the other side, propose that with no prior beliefs, one is not
driven by established ways of thinking or creating, opening the possibilities of break-
through insights where nobody had been before [3].
Research Theoretical Framework. The following theories and methods have
been used in work: the system approach [4], information extraction, exploratory ex-
perimentation [5], the methods of statistical analysis [6], simulation modelling ap-
proach [7], methods of quality analysis [8], surveys, pattern identification [9], build-
ing a model.
Limitations of the research are linked to the research base: one industry for start-
ups - aviation industry, one hundred experiment participants and their level of exper-
tise (undergraduate students, non-native English speakers), task complexity and am-
biguity, ambiguity of natural language.
Scientific novelty:
analysis of Value Proposition in the context of air transportation and startups; con-
struction of Value Proposition corpus / dataset; systematization of Value Proposition
(ontology); analysis of linguistic patterns of Value Proposition; automation of linguis-
tic features detection that could be used to detect Value Proposition (feature engineer-
ing [10])
3 State of the Art
Articles reviewed here and referenced throughout the article were sourced from online
libraries using a relative research approach: papers related to Value Proposition,
startup concept, corpus linguistics, text annotation. The inclusion criteria for an article
in this review were the following: the study was of an exploratory or empirical nature
or gives an idea of the terminology source.
A rich theoretical and empirical literature can be applied to the question on how to
identify Value Proposition in startups texts. As the current research focuses on lin-
guistic patterns of startups landing pages, the author reviews theoretical and empirical
literature on specific online text features, textual value proposition methods, value
proposition comprehension by consumers of different level of expertise.
Michael Lanning and Edward Michaels first used the expression “value proposi-
tion” (VP) in a 1988 work document for the consulting company “McKinsey and Co”.
In the article, which was entitled “Delivering value to customers”, the authors define
value proposition as “a clear, simple statement of the benefits, both tangible and in-
tangible, that the company will provide, along with the approximate price it will
charge each customer segment for those benefits”[11].
In 2016, Eric Almquist suggested a strategy based on a differentiated customer
value proposition. A suggested set of value was called Elements of Value [12]. Ele-
ments of Value categories are based on Maslow’s Hierarchy of Needs shown in Fig-
ure 2.
60
Fig. 2. Heuristic model of value with examples of companies exhibiting elements of value [12]
There are comparatively few researches on analyzing value propositions in online
startups. In 2007, Su-C Li in his paper argues that a properly constructed value propo-
sition is essential to the value creation process in e-business, and value co-production
is the building blocks for value protection mechanism in network economy [13].
Äyväri, Anne, and Annukka Jyrämä in their article “Rethinking value proposition
tools for living labs” published in 2017 provide a conceptual analysis on value propo-
sition tools to be used in future empirical research and in building managerial insight.
The conceptual analysis focuses on a living lab framework and recent theoretical
developments around the concept of value that are reflected in the context of three
managerial tools for creating value propositions. Among findings in the context of the
living labs approach, the Value Proposition Builder seems to conflict with the ideas
and premises of user-centric innovation processes [14]. In 2019, Guo, Hai, Jun Yang,
and Jiaping examined the fit between value proposition innovation and technological
innovation (exploitative vs explorative) for the performance of startups in the digital
environment. They based their research on on-site survey data of 285 digital startups
in one of the world's largest digital economies and found that explorative innovation
strengthens the positive impact of value proposition innovation on the performance of
startups, but exploitative innovation weakens this positive effect [15].
Corpus linguistics has generated a number of research methods, which attempt to
trace a path from data to theory. Wallis and Nelson in 2001first introduced what they
called the 3A perspective - Annotation, Abstraction and Analysis: Annotation consists
of the application of a scheme to texts [16]. Annotations include structural markup,
part-of-speech tagging, parsing, and numerous other representations. Abstraction
consists of the translation (mapping) of terms in the scheme to terms in a theoretically
motivated model or dataset. Abstraction typically includes linguist-directed search
but may include e.g., rule-learning for parsers. Analysis consists of statistically
probing, manipulating and generalizing from the dataset. Analysis might include sta-
tistical evaluations, optimization of rule-bases or knowledge discovery methods.
61
4 Research Design and Preliminary Results
4.1 First Stage – Current
A total of 800 different startups in the field of air transportation were chosen, their
landing pages consist the research base for analysis on Value Proposition. 100 partici-
pants of low level expertise - IT, Aviation, and Management undergraduate students –
were involved in a preliminary survey, post-survey, training and post-training annota-
tion of startups landing pages. A separate webpage was created for conducting sur-
veys, two-level training and annotation process. The quality of annotation as it was
performed by low-expert students was assessed hereafter by industry experts.
The aim of a preliminary survey is to understand the level of expertise for non-
trained participants. The students were told about the basic concept of value proposi-
tion. Also, they were given some examples of Value Proposition such as typical ones:
affordability, quality, speed; less typical: eco-friendliness. After that they were asked
to list as many values as they can, that are provided by the Air Transportation startups
and companies.
Fig. 3. The experiment webpage with training 1 procedure Source: http://ttiv.s3-website-us-
east-1.amazonaws.com/
Figure 3 shows the experiment webpage structure, where participants conduct a
survey, two trainings and an annotation procedure. The Figure reflects the Training 1
procedure. At the end of annotation experiment the participants are asked to conduct a
new survey on Value Proposition provided by the Air Transportation startups and
companies to estimate their level of expertise and compare it with the pre-training and
pre-annotation one. After the participants are explained the startup concept, they
learned existing Value proposition major theories. After that they were offered two
trainings. Training 1 shows five relatively simple examples of how to determine VP
on startups landing pages. Training 2 presents five landing pages where the VP can-
62
not be easily identified due to its vague expression or the creator of the webpage use
non-standard ways to promote their startup, e.g. video or non-trivial words.
The participants have options to identify a landing page as not a startup, can state
that the value is hard to identify, and can name the page as not from the air transporta-
tion industry. Also, they are asked to click a Like button if they think this startup
clearly deliver the VP.
4.2 Second Stage – Future Work
Preliminary results. In the preliminary results the author observed that the current
assumption, based on the suggested algorithm, shows the expected ambiguity of the
natural language as well as the task complexity and ambiguity as perception of the VP
concept depends on the individual personal characteristics and background. Also, it
proves that the necessity of inter-annotator agreement is a measure of how well two
(or more) annotators can make the same annotation decision for a certain category
[17]. This moment 56% of the planned startups landing pages are annotated.
Future Work. The immediately future work is to identify the features (e.g. linguis-
tic, semantic), so the author can determine specific for startups Value Proposition
patterns to develop testable predictions. Creating transportation industry values ontol-
ogy based on inter-annotated corpus with high ambiguity level is one of the objectives
of this research. Building and testing a model is the next phase of the current studies.
As the result of this work the author will elaborate a methodology for analysis of val-
ues based on a model of automatic identification of values in the text of a startup’s
landing page in the air transportation industry.
At the end stage the author plans to develop a tool in the form of software for as-
sisting industry analysts in their research and advisory services.
Factors that influence the performance of the Value Proposition identifying model
are discovered during the annotation process. The author is analyzing participants’
comments and commits not only to identify the clearly viewed linguistic features, but
also to find not so vivid patterns that we can use to automate the process of VP identi-
fication. There are several factors technical and non-technical that the author thinks
can influence the performance of annotation process (like what influences the attitude
towards and the time of annotation) and that are being investigated.
Conclusions
In this research the author empirically has studied the process of Value Proposition
identification in the air transportation startups. The study has been conducted accord-
ing to one of the well-recognized path of corpus linguistic research methods: 3A –
Annotation, Abstraction, Analysis. The current stage of the given research is a landing
text annotation process. An experiment design has been developed: preliminary sur-
vey, two-level training, dataset of objects, 100 annotators.
The preliminary findings show the importance of inter-annotator agreement and an
ambiguity of natural language and value identification process. The next stage of this
research will include identifying the linguistic features and patterns. It will be a basis
63
of elaborating a methodology of automatic identification of values in the text of a
startup’s landing page in the air transportation industry. At the end, the author plans to
develop a tool in the form of software to assist industry analysts in their research and
advisory services.
References
1. Skytran company webpage (online), www.skytran.com, last accessed April 12, 2020
2. McCue ME, McCoy AM. The Scope of Big Data in One Medicine: Unprecedented Oppor-
tunities and Challenges. Front Vet Sci, e-publishing (2017).
3. Shih W, Sen C. Data-Driven vs. Hypothesis-Driven Research: Making Sense of Big Data.
Academy of Management Proceedings (2017).
4. von Bertalanffy, L. (1968). General System Theory: Foundations, Development, Applica-
tions. New York: George Braziller (1968).
5. Burian R.M. Exploratory Experimentation. In: Dubitzky W., Wolkenhauer O., Cho KH.,
Yokota H. (eds) Encyclopedia of Systems Biology. Springer, New York, NY (2013).
6. Ravens C. Methods of statistical analysis. In: Internal Brand Management in an Interna-
tional Context. Innovatives Markenmanagement, Springer Gabler, Wiesbaden (2014).
7. Simulation Modeling. In: Alhajj R., Rokne J. (eds) Encyclopedia of Social Network Anal-
ysis and Mining. Springer, New York, NY (2018).
8. Müller P., Pickard K., Bertsche B. Analysis and Inclusion of Synergies of Common Quali-
ty Management Methods for Optimised Quality Assurance. In: Spitzer C., Schmocker U.,
Dang V.N. Probabilistic Safety Assessment and Management. Springer, London (2004).
9. Fu K.S. Syntactic (Linguistic) Pattern Recognition. In: Fu K.S. Digital Pattern Recogni-
tion. Communication and Cybernetics, vol 10. Springer, Berlin, Heidelberg (1976).
10. Brownlee, Jason. Discover Feature Engineering, How to Engineer Features and How to
Get Good at It, machinelearningmastery. com, (2014)
11. Harvey Golub, Jane Henry, John L. Forbis, Nitin T. Mehta, Michael J. Lanning, Edward
G. Michaels, and Kenichi Ohmae. Delivering value to customers. Harvard Business Re-
view, President and Fellows of Harvard College (1988).
12. Eric Almquist, John Senior, Nicolas Bloch, “The Elements of Value”, Harvard Business
Review, pp. 46–53 (2016)
13. S. Li, "The Role of Value Proposition and Value Co-Production in New Internet Startups:
How New Venture e-Businesses Achieve Competitive Advantage," PICMET '07 - 2007
Portland International Conference on Management of Engineering & Technology, Port-
land, OR, pp. 1126-1132 (2017).
14. Äyväri, Anne; Jyrämä, Annukka: Rethinking. Value proposition tools for living
labs.In: Journal of Service Theory and Practice, Vol. 27, No. 5, p. 1024-1039 (2017)
15. Guo, Hai, Jun Yang, and Jiaping Han. "The Fit Between Value Proposition Innovation and
Technological Innovation in the Digital Environment." IEEE Transactions on Engineering
Management (2019).
16. Wallis, S. and Nelson G. Knowledge discovery in grammatically analysed corpora. Data
Mining and Knowledge Discovery, 5: 307-340.(2001)
17. Artstein R. Inter-annotator Agreement. In: Ide N., Pustejovsky J. Handbook of Linguistic
Annotation. Springer, Dordrecht (2017).
64