Semanticizing Sociability: Documenting Relationships in the context of Cultural Heritage

Tracking #: 2755-3969

Authors: 
Stephen Hart
Karine Léonard Brouillet
Philippe Michon

Responsible editor: 
Special Issue Cultural Heritage 2021

Submission type: 
Full Paper
Abstract: 
Museums are increasingly leveraging the relationships between actors to guide their work and uncover new socio-cultural networks. Better socio-cultural network modelling based on CIDOC CRM—which is not specifically concerned with actor networks but remains the most widely used ontology in the heritage community—would contribute to the discovery, dissemination and enhancement of museum information and would promote inter-institutional collaboration. To achieve this, models must consider complex networks of inter-entity relationships and ways of inferring and representing such relationships. Various ways of representing actors in networks are already available (Bio CRM, Linked.Art, etc.), but none (based on CIDOC CRM) seem to focus on modelling the network itself. This article, after a critical analysis of relationship patterns in relevant models that offer them, proposes a CIDOC CRM-based approach to representing the events that concretize social interactions and relationships. This is done by representing actors’ roles within events, with patterns that are non-hierarchical (one actor does not take precedence over another) and multi- or bi-directional (links between actors are reciprocal) in order to provide a more detailed description of socio-cultural networks.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Major Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Julie M. Birkholz submitted on 26/May/2021
Suggestion:
Minor Revision
Review Comment:

This paper entitled “Semanticizing Sociability: Documenting Relationships in the context of Cultural Heritage” proposes a linked data model for modeling social relations inferred from historical records. The paper is well-written, and clear. It outlines the importance of modeling relations but also the challenges in re-piecing the networks from the past, particularly when relational information is not specifically conserved. The proposed model is original and makes a contribution to the field by extending current models.
Modeling of real social networks is always a challenge, as relationships are often modelled from reported or inferred relationships. For example, a kinship relationship of siblings may be recorded in a birth record in connection to a shared birth mother or father, while relations are also inferred through affiliation- shared attendance at a event or board members of an organization, and so forth. As the authors so clearly explain in addition to not only having scattered records, there is also a lack of a compatible format for information that allows users – historians, curators or the public, to easily identify relations between items. In addition, the ways that networks evolve is dependent on the context. Thus in working to understand networks we not only need information about the entities – nodes, and relations – edges, but also other factors of time, place and attributes of the entities and relations. This makes modeling records and historical information of social networks extremely complex. The authors do a great job of explaining the shortcomings of minimalist approaches for conceptualizing social relations.

Although, I agree with the propositions, and I am very curious about how this model can be implemented in practice; it is notable that few methodological references of networks and network data collection are mentioned. Work on social network analysis as a method and the validity and reliability of conceptualizing relations is an entire sub-field of social networks that could be integrated into this work to further solidify the proposed arguments. I would suggest starting out by reviewing: 1) Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications and 2) Marsden, P. V. (1990). Network data and measurement. Annual review of sociology, 435-463. And specifically for historical social networks: 1) Bearman, Peter S., Moody, J. & Faris, R. (2002). “Networks and History.” Complexity, no. 8: 61–71. And 2) Van De Camp, M., & van den Bosch, A. (2011). A link to the past: constructing historical social networks.
In Proceedings of 2nd Workshop Computational Approaches to Subjectivity & Sentiment Analysis (pp. 61-69).
In addition, I would question the authors’ emphasis on structural parity in the proposed model and as a requirement for asserting or validifying social networks from historical records. I do not disagree that it would provide certainty in inferring networks and it is also a core practice for curators and researchers in confirming or asserting information; although incompleteness of information about social networks both missing nodes and relations is inherent in social networks. See: 1) Kossinets, G. (2006). Effects of missing data in social networks. Social networks, 28(3), 247-268. And 2) Wang, D. J., Shi, X., McFarland, D. A., & Leskovec, J. (2012). Measurement error in network data: A reclassification. Social Networks, 34(4), 396-409. And 3) Borgatti, S. P., Carley, K. M., & Krackhardt, D. (2006). On the robustness of centrality measures under conditions of imperfect data. Social networks, 28(2), 124-136. Instead I would rather suggest to the authors to consider how to implement a confidence interval for relations based on the number or type of confirmed relations. This is also extremely important for valorization in making these data available for research, as there is a drastic difference in what role this interval may play in practice for research. For example, the importance of a confidence interval would be different for a researcher seeking a list of relations to decide on where to do a closer reading, versus a network researcher who would seek to analyze the structure and positions of the historical networks. Given this was a presentation of the proposed model the aspect of valorizing relational data of collections was not introduced, although I look forward to following the authors' future work to see how it is being implemented in practice and how the model will evolve considering the shapes of the data available and context of the networks recorded.

Please note there is something wrong with LaTeX file with figure 11 that cuts that page and thus columns oddly in half.

This paper presents a novel model for modeling relations and is thus useful for cultural heritage institutions and researchers in considering how to model and thus conserve, store and organize historical relational information of their collections. I recommend this paper for publication in this special issue; pending 1) the integration of aspects of social network methodology, and 2) the consideration of a proposal for integrating the provenance and thus a confidence interval for addressing the issue of structural parity, to further strengthen their arguments.

Review #2
By Bruno Sartini submitted on 29/Sep/2021
Suggestion:
Major Revision
Review Comment:

This paper offers a complete survey on the state of the art of social relationships ontologies and patterns, and then defines a new model based on existing CIDOC CRM classes and properties to describe such relationships. However, there are some keys issues in the purpose, evaluations, proposed model, and the conclusions.
Firstly, not many examples of relationship “in the context of Cultural Heritage” are provided. In the paper it is not clearly explained whether there are some specific relationships that only happen in the context of cultural heritage, and this doubt is enforced by the repeated example throughout the paper of the relationship between a thesis supervisor and their student, which resembles a relationship that is not strictly related to cultural heritage domain. There are no real requirements or competency questions through which the authors explain in detail what kind of information needs to be extracted from such model.
Secondly, in the paper it is discussed many times how the verbosity or complexity of potential queries degrades the quality of a model. However, this aspect is not evaluated when the discussion goes to the proposed model. A comparison between the verbosity of queries in the different ontologies and/or patterns would help potential readers to understand the difference in this specific aspect, and whether the proposed model does not have some pitfalls in this aspect compared to the others.
Moreover, the conclusion of the paper seems contradictory compared to some statements made inside it. At many times, inside the paper, it is stressed how the creation of new classes and properties for CIDOC should be avoided for different reasons. But, in the conclusion, this is the direction that seems to be suggested by the authors.
Finally, as admitted by the authors, CIDOC crm was not developed aiming at describing social relationships, its value relies in describing elements belonging to the cultural heritage domain. Given the lack of coverage of CIDOC for the purpose of the paper, and the necessity to still be linked to this model that is a standard for the cultural heritage domain, perhaps ontology that is developed with the aim of this paper and then aligned to CIDOC would be a better solution. This option was not discussed by the authors. Reusing CIDOC, that is considered by the authors itself not valid enough to describe different relationships does not seem to improve the state of the art on the topic.
In summary, the paper does not stress enough what are the differences between standard social relationship and ones you can find in the context of cultural heritage, some evaluations made on the evaluated existing models are not reflected on the proposed model, the conclusion seem to be contradictory with some statements in this work, and the general solution adopted does not improve current SOtA due to the approach chosen (only reusing CIDOC).
A major revision of the work is suggested.

Review #3
By Marieke van Erp submitted on 24/Oct/2021
Suggestion:
Reject
Review Comment:

This manuscript was submitted as 'full paper' and should be reviewed along the usual dimensions for research contributions which include (1) originality, (2) significance of the results, and (3) quality of writing. Please also assess the data file provided by the authors under “Long-term stable URL for resources”. In particular, assess (A) whether the data file is well organized and in particular contains a README file which makes it easy for you to assess the data, (B) whether the provided resources appear to be complete for replication of experiments, and if not, why, (C) whether the chosen repository, if it is not GitHub, Figshare or Zenodo, is appropriate for long-term repository discoverability, and (4) whether the provided data artifacts are complete. Please refer to the reviewer instructions and the FAQ for further information.

The paper presents a modelling exercise to capture relationships between people in a cultural heritage context. While indeed social networks of historical persons can provide context to cultural heritage collections and present a new way of accessing and interpreting collection objects, I am not convinced by the paper in its current state that this is the best way of modelling this. I have two main concerns with the paper:

- It is not clear to my why CIDOC CRM is the best ontology to model this - why not choose a network specific ontology such as FOAF?
- The presented work lacks an evaluation, how do we know that the modelling choices are made are indeed optimal for modelling this knowledge? The paper provides no dataset and associated competency questions, use case, or other metric that provides an indication of the model quality and usefulness. It seems the authors are not entirely sure yet either as section 3 states "Nonetheless, interactions could be treated as events (crm:E5_Event) rather than activities (crm:E7_Activity). This would make it possible to designate actors as participants in the relationship event which would also make it possible to document unintentional social interactions." While the authors note that it would require additional modelling on the participant, it seems more expressive. A targeted evaluation could shed light on whether the additional modelling is worthwhile.

Minor comments:
- Maybe move footnote 8 to the main text?
- Section 2: complexifies -> complicates?
- roles^16, -> roles, ^16
- Figure 10: are temporal boundaries modelled too? e.g. a thesis supervisor role is generally active only for a particular time frame, although a person will always have been someone's thesis supervisor
-

Review #4
Anonymous submitted on 13/Jan/2022
Suggestion:
Major Revision
Review Comment:

(1) Originality: The authors are not able to explain the reason why modeling social interactions is important for the enhancement of museum information or promoting collaboration, as they claim. The examples they give about brothers or wife and husband, and sellers seem (largely?) irrelevant for the cultural heritage domain. Once they decide on tackling this problem, they make a good review of previous approaches but their solution is rather straightforward, ASSUMING that the requirement of structural parity is such a cornerstone requirement in this area. The body of the paper fails to explain this and claims that such structures facilitate reasoning without any kind of explanation nor examples. Additionally, I think they are at times mixing their three guiding principles with the need for n-ary relations and defining some sort of a context.
(2) significance of the results: Very likely useful for the CIDOC CRM ontology, but since I believe the paper didn´t make a convincing case for either the problem nor the technical challenge of structural parity, I don´t see an impact beyond that.
(3) quality of writing: I found it very hard to read and unintuitive. Claims are made without (at least) examples. "That" / "These" / "the latter" / etc are used several times after long paragraphs and it´s not clear which of the many previous concepts they refer to. Notations in the figures are sometimes confusing (ellipses vs rectangles). The authors talk about intentionality and evolving vs one-time characteristics of activities; E7_Activity indeed implies intentionality, but seems to include long-lasting actions. Reproducing the definition they refer to would have been clearer and shorter. Lastly, there are several phrases whose meaning I don´t understand at all, for instance on page 13, "contains a potential for precision [...]". Precision of the strength of the relationship (but there are no probabilities associated, right?) and if not, then precision of what?