Identifying Semantic Frames in Adventure Tourism (Short Paper) Isabel Durán-Muñoz† and Eva Lucía Jiménez-Navarro∗,† 1 Universidad de Córdoba, Plaza Cardenal Salazar, s/n, 14003 Córdoba, Spain Abstract This paper proposes a methodological approach to the identification of semantic frames (1, 2, 3) in the language of adventure tourism. This method is mirrored on the existing entries recorded in the specialized resource DicoAdventure, an online bilingual (English, Spanish) dictionary about adventure tourism. More specifically, annotated contexts and argument structures of motion verbs were compared to the descriptions and the frame elements of the entries collected in FrameNet insofar as it served as a reference point. A total of 13 semantic frames were discovered and most of them corresponded to frames already described in this database; however, some new ones also emerged. Additionally, some units were not included in FrameNet or, when they were, they evoked different frames depending on whether they referred to the general language (FrameNet) or the specialized language (DicoAdventure). Keywords adventure tourism, annotated context, argument structure, frame element, motion verb, semantic frame 1 1. Introduction During the past two decades, Frame Semantics (1, 2, 3) and its application FrameNet (henceforth, FN) has drawn the attention of an increasing number of terminologists interested in developing specialized resources that represent the connection of specialized units with semantic frames. According to L’Homme et al. (4), semantic frames are especially attractive in terminology, since it is assumed that there is a connection between the conceptual structure of specialized fields of knowledge and the linguistic units used to convey this knowledge. Compilers of domain-specific resources have either adhered closely to the methodology developed within the FN project or adapted it to achieve different objectives. Most recent examples of specialized domains for which resources based largely or loosely on Frame Semantics are aeronautics, linguistics, the environment, and sport, among others (cf. 4, 5, 6, 7, 8). In this paper, we propose a method for identifying semantic frames in the specialized domain of adventure tourism, an increasingly popular tourism segment which combines nature, risk, and adventure in a range of activities performed by tourists. This methodology follows ―with some adjustments― the Frame Semantics principles and the FN methodology and is mirrored 3rd International Conference on “Multilingual digital terminology today. Design, representation formats and management systems” (MDTT) 2024, June 27–28, 2024, Granada, Spain ∗ Corresponding author. † These authors contributed equally. iduran@uco.es (I. Durán-Muñoz); lucia.jimenez@uco.es (E. L. Jiménez-Navarro) 0000-0002-6795-498X (I. Durán-Muñoz); 0000-0001-9377-6921 (E. L. Jiménez-Navarro) © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR ceur-ws.org Workshop ISSN 1613-0073 Proceedings on existing entries of motion verbs included in DicoAdventure, a bilingual (English, Spanish) specialized dictionary about adventure tourism. Taking into account the findings reported in previous studies (9, 10) as well as the assumption that frames are particularly useful when analyzing predicative terms (11), we focused on motion verbs 2 extracted from the ADVENCOR corpus, a specialized corpus about adventure tourism (12), and included in the DicoAdventure dictionary (cf. Section 2), to reach the goals of this study, which are: on the one hand, to discover the semantic frames that are evoked by motion verbs (representing either real or fictive motion) in this specific domain, and, on the other, to identify the different terms in English that are linked to those frames. As in previous studies (11), the source of information employed in DicoAdventure was not aimed at identifying frames and the terms that evoke them, but it is proved that these frames can be discovered in a subsequent phase according to the lexico-semantic properties of specialized units. To do so, this work is structured as follows. Section 2 briefly describes the DicoAdventure resource, the source of information employed to identify the semantic frames evoked by motion verbs in adventure tourism. Section 3 presents the methodological approach, along with a description of the data that was used, and Section 4 discusses the most relevant results. Finally, we provide some concluding remarks and mention a few directions for future work in Section 5. 2. The DicoAdventure dictionary The DicoAdventure dictionary is an online bilingual (English-Spanish) lexico-semantic specialized resource about the language used in adventure tourism freely available at http://olst.ling.umontreal.ca/dicoadventure/ and still under construction. The primary focus of the resource has been on motion verbs, representing both real and fictive motion, after having noticed that they overwhelmingly predominate in this domain (10, 13, 14). Besides, other types of concepts have also been recently included, especially nouns referring to adventure activities, such as trekking, kayaking, canyoning, or climbing. All the entries gathered in the resource offer the same kind of information, which is divided into linguistic, pragmatic, and semantic, and can be easily visualized. Figure 1 below shows the entry for abseil1 and contains: (1) the term itself (it comes with a number because there is an entry for every meaning conveyed; for example, we can find the entries ascend1 and ascend2 in which real motion and fictive motion are represented, respectively) together with its grammatical category, (2) definition and argument structure of the term, (3) linguistic realizations of the arguments and examples, (4) equivalents in the other language, (5) contexts of use, annotated contexts, and a summary of the annotations, and (6) conceptual relations with other terms (e.g., synonyms, collocations, different parts of speech, etc.). Additionally, some administrative information is included on the upper right corner of the entry together with a link to the top of the page, along with a picture which represents the definition of the term and facilitates the interpretation of its meaning. 2 In this context, we consider motion verbs as “verbs describing a displacement of an entity, either a person or a thing, in space” (10). Figure 1: Entry for abseil1 in DicoAdventure For the purpose of this study, the most relevant part in this resource is the argument structure and the annotated contexts, which provide us with relevant information to discover and define the semantic frames embodied in adventure tourism; moreover, it allows us to identify the different terms belonging to these frames. The methodology followed is explained in the following section. 3. Methodology To identify the semantic frames that were evoked by the motion verbs included in DicoAdventure we took the following steps: 1. We checked the terms in FN and considered the semantic frames assigned to them. For example, head-v is associated with SELF_MOTION and LEADERSHIP (cf. Figure 2). Figure 2: Semantic frames evoked by head-v in FN 2. We carried out an analysis of the description of the frames and the distinct elements they included ―which can be classified as core (obligatory participants) or non-core (non-obligatory) participants―, and selected the most convenient for our purpose. For the case of head-v in adventure tourism, the corresponding frame was SELF_MOTION. 3. We analyzed the argument structure and the annotated contexts of the occurrences of the verbs in the DicoAdventure resource and concluded whether the semantic frames identified in FN could also be used in our specific domain or, on the contrary, a new frame was required. If the former situation arose, the frame was used and its name adapted (if necessary); this is the case of the frame SELF_MOTION. In contrast, if there was the need to create a new frame, this was proposed and named according to the information extracted from DicoAdventure, such as the frame MOTION_FROM_SOURCE, which includes motion verbs with a focus on the origin of the motion and does not exist in FN. Along with the identification of frames, we assessed the different motion verbs that evoked the same semantic frame according to the following criteria established by the FN methodology (15): • The units should have the same number of arguments (both explicit and implicit). • The units should have the same type of arguments (both explicit and implicit). • The units should denote the same part of scene. • The units should have the same relations, both within the frame elements and with background frames. • The units should have the same frame element profiling. As can be observed, contrary to the FN methodology, here the semantic frames (as well as the terms evoking them) were discovered after the terms were examined in the annotated contexts in DicoAdventure, instead of being postulated prior to the descriptive work. Besides, labels used for most obligatory elements in the semantic frames were different in both resources, since DicoAdventure employs specific labels for the adventure tourism language, such as TOURIST or DESTINATION. Nevertheless, we considered that the labels were equivalent if their referent was the same. Thus, AGENT was considered to correspond to TOURIST in DicoAdventure, DESTINATION to GOAL, and AREA to PLACE. The following section provides the most relevant findings after the implementation of this methodology. 4. Semantic frames in DicoAdventure As mentioned above, the information collected in the FN database served as a reference point to define the semantic frames evoked by the motion verbs included in DicoAdventure. This database provided a robust foundation for understanding the meanings of verbs, enhancing comprehension of how knowledge is structured within adventure tourism. However, it is crucial to note that, while this resource was valuable, it could not address all the needs identified in studying adventure tourism’s specialized discourse, as FN represents the general language. Consequently, some tailored frames were established to conduct a more comprehensive analysis and to meet the specific requirements of this specialized language. As a way of example, we provide the description of the semantic frame SELF_MOTION in FN (cf. Figure 3) and the argument structure of head1 in DicoAdventure (cf. Figure 4), since, according to FN, the verb head-v evokes this frame. Following the FN proposal, the same name of the frame is also used in DicoAdventure. Nonetheless, when we look closer at the obligatory elements in the units, differences regarding the name of the semantic roles are found because of the specialization of the verb in DicoAdventure, that is, some elements are named differently, such as SELF_MOVER in FN and TOURIST in DicoAdventure, AREA in FN and PLACE in DicoAdventure, or GOAL in FN and DESTINATION in DicoAdventure. Therefore, even though the same frame can be used in the adventure tourism resource, some adjustments are required to match the specifications of the specialized domain. Figure 3: Description of the semantic frame SELF_MOTION in FN Figure 4: Argument structure of head1 in DicoAdventure After the implementation of the described methodology, a total of 13 semantic frames were unearthed (cf. Table 1). Table 1 Semantic frames evoked by the motion verbs in DicoAdventure (current version) ARRIVING MOTION_DIRECTIONAL SELF_MOTION CAUSE_MOTION MOTION_FROM_SOURCE SELF_MOTION_WITH_SAFETY_INSTRUMENT CAUSE_TO_LAND OPERATE_VEHICLE TRAVERSING DEPARTING PATH_SHAPE MOTION RIDE_VEHICLE Most of these frames were found in FN and their designation was taken from it. Only two showed no correspondence in the resource and new semantic frames were suggested for the domain of adventure tourism, namely MOTION_FROM_SOURCE and SELF_MOTION_WITH_SAFETY_INSTRUMENT. For their part, when looking at the units that evoke those frames (cf. Appendix), three different situations were observed: 1. The units were exactly the same and evoked the same frame, such as in ARRIVING or DEPARTING, although the obligatory elements were named differently (e.g., GOAL in FN and DESTINATION in DicoAdventure). 2. The units evoked different frames in each resource, such as climb-v, disembark-v, or slide-v. 3. Some units were not included in a semantic frame available in FN, such as kayak-v or skydive-v. Table 2 shows the number of units that represent these three different situations. Table 2 Analysis of results Category Number of units Same frame 37 Different frame 11 No frame in FN 9 TOTAL 57 The analysis reveals interesting findings about how situations that would appear to be similar can be conceptualized differently in everyday language (as represented in FN) and in specialized situations (as evidenced by the data collected in DicoAdventure), such as the facts that most of the identified frames are found in FN despite having different frame elements or that new frames are required to be created. As an illustration, paddle-v is categorized within the semantic frame OPERATE_VEHICLE in both FN and DicoAdventure. Nonetheless, despite belonging to the same frame, we must point out the different uses of the verb in every resource. Considering the information provided in FN, the words gathered in this frame involve motion that requires a vehicle and someone who controls it. One of the core elements highlighted is PATH, which refers to the trajectory of motion and includes directional expressions and “middle of path” expressions, encapsulated in constructions where the verb functions as an intransitive verb. Up to here, the notion of OPERATE_VEHICLE in DicoAdventure is exactly the same. However, regarding the frame element PATH, specific contexts belonging to the domain of adventure tourism show that paddle-v can also be used transitively when this is followed by a path (demonstrating the special behavior of this verb in the specialized discourse), such as You can experience unforgettable outdoor activities, scale heights, paddle the river and explore underground [ADVENCOR]. For its part, the only transitive uses of this verb in the general language, as depicted in both FN and the definition of this verbal unit in the Collins Dictionary, 3 only includes the vehicle that is operated in the direct object position (e.g., Tim drove his car all the way across North America [FN], just like DicoAdventure also does (e.g., … then navigate a course around obstacles or paddle your raft to an island for a picnic! [ADVENCOR]). Another key point is the fact that the same unit can evoke different frames in the two types of language, that is, general (FN) or specialized (DicoAdventure). Thus, disembark-v is included 3 https://www.collinsdictionary.com/dictionary/english/paddle [Last accessed: 16/03/2024). in the semantic frame DISEMBARKING in FN, which is defined as “A TRAVELLER leaves from or dismounts a VEHICLE”. Comparatively, this definition is the one that we hold after exploring the annotated contexts gathered in DicoAdventure. Nevertheless, our main focus in the specialized domain under study is on the starting point of a motion as emphasized in the definition itself, so we consider that this verb fits better into the abovementioned ad hoc frame MOTION_FROM_SOURCE. Two other verbs belong to this frame too, namely jump-v and launch- v, both of which evoke a distinct frame in FN, to know, SELF_MOTION and CAUSE_MOTION, respectively. A similar case is the verb slide-v, included in the semantic frame MOTION in FN, it being defined as “Some entity (THEME) starts out in one place (SOURCE) and ends up in some other place (GOAL), having covered some space between the two (PATH)”. However, from this definition we can infer that THEME (i.e., the entity moving) is the most prominent element in the situation, therefore, we can safely assume that categorizing slide-v as a unit evoking the semantic frame SELF_MOTION is perfectly suitable in the context of adventure tourism. In fact, FN defines this frame as “The SELF_MOVER, a living being, moves under its own direction along a PATH”, which is precisely what specialized contexts show, for instance, Book canyoning if you like the idea of jumping off the rocks and sliding down the waterfalls [ADVENCOR]. This doubleness seems to emerge from the distinction between a non-living entity and a human being moving themselves, conceived of in FN. Finally, we must also mention those specialized units that were not collected in FN. A couple of examples are skydive-v 4 and navigate-v, which are included in the SELF_MOTION and OPERATE_VEHICLE frames in DicoAdventure, respectively. Again, this categorization means that the emphasis in skydive-v is on the entity performing the motion (which is at the same time the one moving; e.g., At the top of the South Island you can skydive over the Abel Tasman national park [ADVENCOR]), on the one hand, and on the vehicle needed when the motion takes place, reflected in navigate-v, on the other. As for navigate-v, unlike the case of paddle-v previously mentioned, the general language also makes use of this verb as a transitive one with a path in the direct object position as well as the vehicle itself. 5 It also occurs in the specialized domain of adventure tourism, as represented in the following examples extracted from ADVENCOR: Feel your adrenaline pumping as you navigate powerful rapids and white water; … you will find lots of people who know how to navigate a canoe …. Nevertheless, in this specialized language the presence of a path is stronger than the presence of a vehicle after the verb in the direct object position, as observed in the contexts retrieved from the corpus (cf. Figure 5). 4 In fact, skydive-v was shown to enjoy the highest keyness score in the language of adventure tourism, that is, the most typical one in this specialized discourse (16). 5 https://www.collinsdictionary.com/dictionary/english/navigate (second meaning provided). [Last accessed: 16/03/2024). Figure 5: Sample of contexts containing navigate-v in ADVENCOR 5. Conclusions This work proposed a bottom-up methodology to discover semantic frames in a specialized domain, adventure tourism, taking the English version of FN as a reference point. After a thorough analysis of 57 motion verbs in the language of adventure tourism in English following the described methodology, 13 semantic frames were identified. The results show that most of the motion verbs studied (37) could be assigned to existing frames in FN, such as SELF_MOTION, ARRIVING, or DEPARTING; for others (9), new frames were suggested, for instance, SELF_MOTION_WITH_SAFETY_INSTRUMENT. Furthermore, some units (11) were required to be re-allocated, that is, they evoked different semantic frames depending on whether they referred to the general language (in FN) or the specialized language (in DicoAdventure), like disembark-v or slide-v. The underlying motivation for re-allocating these verbs was the focus on the definition of the frames, that is, the starting point of the motion in disembark-v and the entity moving in slide-v, regardless of their humanness status. Finally, there were some units that were not included in FN because of their specialized meaning, such as skydive-v or navigate-v, and, therefore, we were not able to compare the frames evoked. All in all, the DicoAdventure dictionary is still under construction and this will lead to the discovery of additional frames in the future, some of which will correspond to existing frames, some of which will be new. Added to that, relations between frames already identified can be perceived intuitively (e.g., SELF_MOTION and SELF_MOTION_WITH_SAFETY_INSTRUMENT), but we will further need to investigate this. As to the advantages offered by the methodological approach explained in this paper, we believe that it can be of great help to terminologists and terminology work insofar as it contributes to the distinction between verbs that represent concepts, that is, meanings conveyed in specialized languages, from verbs and their meanings in the general language, as shown with the uses of the verb paddle-v in Section 4. Last but not least, it is worth mentioning some of the implications of this study. First, one of them is connected with the Spanish language, as this is also included in DicoAdventure at present. So far, the method shown in the previous pages has only been applied to the analysis of the entries collected in DicoAdventure in the English language, so the frames evoked by Spanish verbs are also to be explored and compared to the English results. Second, one more implication might be to employ this methodology to study other specialized domains that have not been covered yet, so we hope it can inspire other similar investigations in the future. Acknowledgements This work has been partially carried out within the framework of the R&D project VIP II (PID2020-112818GB-I00) and Recover (Ref. ProyExcel_00540). References [1] C. J. Fillmore, Frames and the semantics of understanding, Quaderni di Semantica 6(2) (1985) 222–254. [2] C. J. Fillmore, C. R. Johnson, M. Petruck, Background to FrameNet. International Journal of Lexicography 16(3) (2010) 235–250. [3] C. J. Fillmore, C. Baker, A frames approach to semantic analysis, in: B. Heine, H. Narrog (Eds.), The Oxford Handbook of Linguistic Analysis, OUP, Oxford, 2010, pp. 313–339. [4] M.-C. L’Homme, B. Robichaud, C. Subirats, Building multilingual specialized resources based on FrameNet: Application to the field of the environment, in: T. Torrent, C. F. Baker, O. Czulo, K. Ohara, M. R. L. Petruck (Eds.). International FrameNet Workshop 2020. Towards a Global, Multilingual FrameNet. Proceedings, Workshop of the Language Resources and Evaluation, LREC, 2020, pp. 85–92. [5] A. Ostroški Anić, Generic semantic frames and frame elements in AirFrame (poster), Proceedings of EAFT Terminology Summit 2018. 3M4Q: Making, Measuring, Managing Terminology. In the Pursuit of Quality (2018), San Sebastian, Spain. [6] P. Malm, S. Mumtaz Virk, L. Borin, A. Savera, LingFN. Towards a FrameNet for the linguistics domain, in: T. Timponi Torrent, L. Borin, C. Baker (Eds.), International FrameNet Workshop. 2018. Multilingual FrameNets and Constructions. Proceedings. Language Resources and Evaluation (LREC 2018), Miyazaki, Japan, pp. 7–43. [7] R. Chishman, B. da Silva, A. Nardes dos Santos, A. L. T. Vianna, S. de Oliveira, M. L. Martins, G. M. de Schryver, Building a paralympic, frame-based dictionary – Towards an inclusive design for Dicionário Paraolímpico (Unisinos/Brazil), in: Proceedings of the EURALEX XIX, Conference of the European Association for Lexicography, Lexicography for Inclusion. Virtual, Book Volume 2, 2021, pp. 723–731. [8] M. T. Ortego-Antón, e-DriMe: A Spanish-English frame-based e-dictionary about dried meats. Terminology 27(2) (2021) 294–321. doi: 10.1075/term.20013.ort. [9] I. Durán-Muñoz, DicoAdventure y la terminología del turismo de aventura: Propuesta de diccionario en línea, in: T. Barceló Martínez, I. Delgado Pugés, F. García Luque (Eds.), Tendencias Actuales en Traducción Especializada, Traducción Audiovisual y Accesibilidad, Tirant Lo Blanch, Valencia, 2021, pp. 395–417. [10] I. Durán-Muñoz, M.-C. L’Homme, Diving into English motion verbs from a lexico-semantic approach. A corpus-based analysis of adventure tourism. Terminology 26(1) (2020) 33–59. doi: 10.1075/term.00041.dur. [11] M.-C. L’Homme, B. Robichaud, C. Subirats. Discovering frames in specialized domains. Language Resources and Evaluation, LREC 2014, Reykjavik (Island). [12] I. Durán-Muñoz, E. L. Jiménez-Navarro, Colocaciones verbales en el turismo de aventura: Estudio contrastivo inglés-español, in: G. Corpas Pastor, M. R. Bautista Zambrana, C. M. Hidalgo-Ternero (Eds.), Sistemas Fraseológicos en Contraste: Enfoques Computacionales y de Corpus, Comares, Granada, 2021, pp. 121–142. [13] I. Durán-Muñoz, E. L. Jiménez-Navarro, Motion verbs in adventure tourism: A lexico- semantic approach to fictive meaning. International Journal of English Studies 23(1) (2023) 27–48. doi: 10.6018/ijes.532851. [14] E. L. Jiménez-Navarro, I. Durán-Muñoz, Collocations of fictive motion verbs in adventure tourism: A corpus-based study of the English language. Revista Española de Lingüística Aplicada/Spanish Journal of Applied Linguistics (2024) 1–25. doi: 10.1075/resla.21042.jim. [15] J. Ruppenhofer, M. Ellsworth, M. Petruck, C. Johnson, C. Baker, J. Scheffczyk, FrameNet II: Extended Theory and Practice (2016). URL: https://framenet.icsi.berkeley.edu/fndrupal/index.php?q=the_book (Last access: 10/01/2024). [16] E. L. Jiménez-Navarro, Treatment and Representation of Verb Collocations in the Specialized Language of Adventure Tourism, Ph.D. thesis, Universidad de Córdoba, Córdoba, 2020. A. Appendix Semantic frame Motion verb Semantic frame Motion verb 1. OPERATE_VEHICLE bike1-v scramble1-v drive1-v skydive1-v fly1-v slide1-v glide1-v swim1-v navigate1-v trek1-v paddle1-v venture1-v ride1-v walk1-v kayak1-v 3. PATH_SHAPE ascend2-v raft1-v climb2-v ski1-v cross2-v surf1-v descend2-v 2. SELF_MOTION dive1-v head2-v head1-v lead1-v hike1-v leave2_v Semantic frame Motion verb Semantic frame Motion verb run2-v soar1-v 4. ARRIVING arrive1-v 8. TRAVERSING cross1-v enter1-v pass1-v approach1-v traverse1-v reach1-v 9. CAUSE_TO_LAND land1-v return1-v 10. DEPARTING exit1-v 5. MOTION_DIRECTIONAL ascend1-v depart1-v descend1-v leave1_v fall1,2-v 11. MOTION_FROM_SOURCE jump1-v 6. SELF_MOTION_WITH_SA abseil1-v disembark1-v FETY_INSTRUMENT climb1-v launch1-v rappel1-v 12. RIDE_VEHICLE ride2-v scale1-v zip1-v 7. MOTION float1-v 13. CAUSE_MOTION pull1-v move1-v