Current travel planning applications provide only a limited support to users for an informed selection of travel solutions. The categorization of travel o ers contributes to solve this limitation complementing basic information and creating awareness on the selection. This paper reports the work done for the conceptualization of o er categories and proposes a solution for enabling their usage in multimodal travel shopping.
departure/arrival and additional search options. The PA sends the generated mobility request to the Travel Solution Aggregator (TSA) that is responsible for providing a set of travel solutions by invoking multiple Travel Experts (TE). Each TE, managed by a TSP or by an intermediary selling TSP solutions, may return zero, one or more itinerary o er items, each de ned as the relation between three concepts: a travel episode (i.e., the segment of the itinerary supported by the TSP), the product, and the passenger(s) involved. After receiving these itinerary o er items from the di erent TEs, the TSA composes them in a set of travel solutions that are then displayed on the user's PA in the form of trips and o ers. Trips de ned by the TSA are bound to one or more related o ers (e.g., rst class and economy o ers for the same trip). The described ow can be enhanced by an additional component de ned in Ride2Rail, the O er Categorizer : an external service used by the TSA to determine the o er categories associated with each o er before sending them to the PA.
On the basis of the presented data model, the proposed conceptualization aims at improving the user awareness in choosing among heterogeneous travel solutions by de ning a shared semantic for their categorization. The next sections describe the analysis of the state-of-the-art on the de nition of o er categories, the conceptualization and the catalogue produced by Ride2Rail, the nal remarks and the next steps. 2
From the analysis of the state-of-the-art, the following patterns to support the
Ride2Rail conceptualization of o er categories emerge:
{ Types of Variables: contributions (e.g., [
Integrating the models proposed in [
The performed analysis of the state-of-the-art highlighted a set of actual variables, belonging to the identi ed macro-areas. However, while instrumental variables are objective and easily measurable, the same does not hold true for perception and symbolic variables. The di culty of obtaining an unambiguous de nition of the concepts makes these factors harder to be appropriately de ned. Nevertheless, an objective quanti cation of perception variables could be evaluated through feedback collected from an adequate statistical sample of users, e.g., measuring the feeling of personal safety or the level of comfort. Symbolic variables are more subjective and, for this reason, cannot be considered to objectively characterize o ers. Therefore, instrumental and perception variables represent potential determinant factors for an o er category, i.e., the variables that can be used to determine the membership of an o er to an o er category.
From the analysed state-of-the-art, we selected a list of actual variables
describing an o er that can be used as determinant factors of low-level o er
categories, i.e., classes that can be easily associated with an o er given its objective
characteristics. It is important to notice that an o er category should be assigned
relatively to the set of o ers provided for a mobility request. For example, the
total travel time variable is the determinant factor for a low-level o er category
that minimizes the total travel time and identi es the quickest travel solution
among the ones available. The complete list of low-level O er Categories
identied and the related determinant factors is reported in [
The assignment of speci c categories identi ed in the state-of-the-art
analysis, like comfortable or environmentally sustainable is not straightforward and
not possible using a single variable describing an o er. For this reason,
contributions from [
The identi ed patterns have been used to support the conceptualization of
the term o er category, while the preliminary analysis of variables and
lowlevel o er categories have been used to de ne a rst catalogue of concrete o er
categories for multimodal travel o ers (brie y described in Section 3 and fully
reported in [
In this section, we report the conceptualization of the O er Category term and its related concepts.
{ O er Feature: An o er is described by a set of objective variables (such as
transportation mode, level of CO2 emissions, cost, etc.). The values assigned
to the objective variables for a speci c o er identify its o er features. For
example, <transportation mode=train> can be a feature of an o er. An
o er feature can be computed considering data provided by the TSP (e.g.,
the price), and/or additional data, such as information related to the trip(s)
associated with the o er in a travel solution (e.g., length in km), or to the
vehicle used in the o er (e.g., CO2 emissions).
{ O er Category: it identi es a set of o ers having speci c shared
characteristics. An o er is assigned to a given o er category considering a set of
o er features, namely the determinant factors for that o er category. The
membership of an o er to a given o er category is de ned by a Category
Score (CS) in the range of [
A rst catalogue of o er categories has been extracted from the state-ofthe-art analysis and subsequently framed considering the provided de nitions. The completeness of the catalogue and the interest in each proposed category have been tested though a survey lled by 609 European travelers3. Rather than computing an exhaustive list of all the possible o er categories, the goal of this catalogue is to elicit the ones that resulted from the survey as the most relevant to provide a comprehensive clusterization of travel solutions obtained in response to a mobility request. In the following, we report the nal Ride2Rail catalogue of o er categories describing the most relevant determinant factors for each of them. The o er categories are ranked according to the relevance attributed by the respondents of the survey. New o er categories, elaborated thanks to suggestions collected though the survey (i.e., Panoramic and Healthy ), are mentioned at the end of the list since we cannot provide an estimation of the relative relevance for the travellers.
{ QUICK: The quick category measures how convenient and e cient the
solution is in terms of time-related issues, considering the total travel time,
the waiting time between legs and the number of stops required. If the
solution includes a segment on-road (e.g., bus/car) and real-time data on tra c
congestion is available, it can be taken into account.
3 The complete description of the survey and its results are available in [
The state-of-the-art analysis has supported the identi cation of a set of patterns considered for the conceptualization of the term O er Category: (i) the assignment of an o er to a given category should consider the speci c values assumed by each variable that describes the o er and the associated trip; (ii) o er categories should be assigned considering a set of objective variables of the o er and should not be conditioned by the characteristics and preferences of a speci c user; (iii) o er categories can be de ned considering multiple variables of the o er. The described conceptualization and the provided catalogue of o er categories is guiding the next step of the Ride2Rail project: the implementation of an o er categorizer component and its testing in four demo sites.
As future work, the de nition of a vocabulary (thesaurus and/or ontology) for o er categories based on the proposed conceptualization would support the provision of enriched interoperable descriptions of travel solutions. Acknowledgments The presented research was supported by the RIDE2RAIL project (Grant Agreement 881825), co-funded by the European Commission under the Horizon 2020 Framework Programme.