Knowledge generation processes are traditionally related to the DIKW (data-information-knowledge-wisdom) hierarchy, a layered model for the classification of human understanding. Software components can be situated in one or several of these layers, or assist in the interfaces between two of them. Most of the knowledge generation processes that occur in the archaeology field involve complex mechanisms of abstraction, relation and interpretation. Is it possible to assist the users in performing these processes? We have detected problems in the archaeological knowledge generation process that could be improved through software assistance. We propose a conceptual framework based on the structure of the data that is being managed by the user, and on the cognitive processes that the user wishes to perform on the data. The proposed framework can, arguably, set the foundation for assisted knowledge generation implemented as software systems.
Archaeology is a vast domain that produces a large amount of data in different
formats: 2D and 3D images, documents, audio, video, etc. and with specific work
methodologies to record and manage data. Archaeology, like other disciplines in
humanities and social sciences, presents complex needs with regard to temporality and
subjectivity[
The strongly human-centered nature of the domain – trying to re-build past events through existing data – focuses these needs in assisting to professionals in knowledge generation. This complex process is the central axis and causes the rest of tasks: conservation, preservation, museum activities, etc. However, the existing support for knowledge generation in this domain is limited to databases applications and some ad hoc approaches to specific data types. There is not an integral solution to assist in knowledge generation processes.
The final goal of the current doctoral research is to provide a conceptual framework that set the foundation for assisted knowledge generation software in archaeological domain, help the users in their cognitive processes: direct observation of the data, complex visualizations, abstraction, etc. This paper summarizes the author’s PhD work and project, working for one year and a half, under the supervision of Dr. Cesar Gonzalez-Perez (Incipit, CSIC) and Prof. Oscar Pastor Lopez (Universitat Politècnica de València). The first six months have been used for studying the domain in a deep way. 2
In this section, the scope of the research is defined: a formal specification of the problems detected and research question formulated. Also, the research methodology followed is discussed. 2.1
Archaeological research builds knowledge based on data about past events. These data are presented in different formats and have been studied for years trying to find special characteristics and improving the recording and management efficiency.
Existing studies identifies specific issues in the archaeological data such
temporality or subjectivity and proposes solutions to support them[
However, building archaeological knowledge is a more complex process than supporting these data issues: it is necessary to study what the archaeologist want to do, what questions asked to the data and how archaeologist reach the initial goals.
Therefore, it is necessary to know the archaeologist’s processes in the generation knowledge from existing data to valuable archaeological knowledge.
This research tries to discover if it is possible to assist the archaeologists in the knowledge generation processes through software. To carry out it and situate the problem statement, a previous study has been required to detect problems and weak points in the archaeological knowledge generation process.
During five months, archaeologists have been observed in a real work environment and have been interviewed about their work methodologies and software use. In addition, a set of questionnaires have been elaborated in order to study deeply the problems found during the observation period. All surveys have been completed by archaeologists (Incipit staff and non-Incipit staff). These professionals have several personal profiles in terms of age, gender and institutional affiliation (public/private institution, educational/non-educational institution).
The following problems in the knowledge generation process have been detected: • Intentional use of uncertainty in the intermediate reasoning to generate knowledge.
This uncertainty is not supported by existing software tools. • Use of reasoning based on geographic and temporal data as a start point in the knowledge generation process. However, this initial reasoning is diffused along the rest of processes. This situation generates confusion about the context of the data in each moment of the knowledge generation process. • Questions asked to the data have not been tracked and specified. This situation involves that non-asked questions are unknown and could be introduce some kind of bias in the knowledge finally generated. • Previous point involves, in addition, that there are no chances to share the asked and non-asked questions between researchers or users working on the same data sources, complicating teamwork. There is no support for the collective generation of knowledge. • Lack of priority management of the different questions asked to the data. (Different level of importance has been detected in the user processes). • Lack of abstract view of the structure of the information management in the knowledge generation process. This problem could form the basis of a low use of the feedback mechanism to build and ask new questions to the data based on the responses obtained in a previous step. • Homogeneous procedures applied to reasoning derived from direct observation and reasoning derived for more complex mechanisms (relation between data, abstraction, interpretation, etc.). This situation could be including confusion about the level of the DIKW hierarchy where the reasoning is situated and the level of subjectivity and uncertainty that is managed.
The problems detected make up the problematic context of the research and the gap founded that this research tries to fill. Our initial hypothesis is that we can improve the knowledge generation process in the archaeological domain (minimizing or reducing the problems detected) by building software models that reproduce and incorporate cognitive facets and needs of the user, and allowing the application of visualization techniques and data-pattern recognition specifically adapted to archaeologists’ ways of working. 2.2
Comparative and empirical studies have carried out to understand the
characteristics of the user, the domain and existing software solutions[
However, there are theoretical models about human knowledge generation that defines the corpus of this research. All existing models follow a hierarchical structure based on layers, with other differences:
Cleveland[
Ackoff[
Carpenter & Cannady[
All compared models fit in with the cognitive character of the processes that allow
us to go up to the next level in the hierarchy. Thus, the characterization of these
cognitive processes into the archaeological domain could be an initial point to solve the
problems detected. Gardin[
Making a complete review of the literature that supports the current research, we
have reviewed the existing methods to assist through software in knowledge
generation processes in other disciplines or domains. The main example founds is the
biomedical domain, in which context Chen[
Chen et al.[
Chen references DIKW[
Regarding TAR[
Category 1: Questions that seek an explanation for a real-world phenomenon, and try to answer what, why or how said phenomenon occurs.
Category 2: Questions that try to build some artifacts from scratch or to improve some existing solution for a problem previously detected.
The main research question of this work emerges from the initial hypothesis. It formulation is: To what extent is it possible to improve the knowledge generation process in archaeology by assisting the user through software tools? In order to ask this research question, the research general goal is to test the initial hypothesis through: • Searching for evidence those support the hypothesis, through user testing and empirical studies. • Searching for evidence against the initial hypothesis, through empirical studies, user observation, validation test and expert feedback.
In the process of answering these questions, additional collateral questions are emerging that allow us to establish a research context. Currently, the research is focused on the secondary question: What are the major existing problems in the knowledge generation process in archaeology? In order to answer the secondary research question, we have defined to detect these existing problems as a specific goal in the first part of the research. 2.4
Our research follows an established scientific methodology approach, based on an initial hypothesis and a general research question to answer. Secondary questions are emerging along the research process. Each secondary research question involves a set of valuable objectives whose commitment level reflects the level of support of the initial hypothesis. In our case, the research questions and the objectives have been exposed in the previous section.
In addition to the methodology general plan following research methods of validation a specific methodology for the requirements elicitation and the study of knowledge generation methods in this domain is proposed.
The interdisciplinary character of the research and user characteristics has been necessary to apply user-centered methods, in order to understand how users build knowledge from the data and how kind of software assistance would be better to support these processes.
On the other hand, the fact that software systems are involved suggests that we need to formalize user needs into requirements.
In this scenario, a hybrid methodology is proposed. All requirement elicitation processes are extracted following shadowing studies, interviews, surveys and user experiments. This allows us to directly observe the users in a real work environment, together with archaeological sets of data and case studies in their domain. In order to contrast the empirical results in an analytical way, other analytical techniques to extract requirements are applied concurrently, such as discourse analysis and datamining. 3
Our proposal involves a conceptual framework that captures and implements the needs of the users who work in the archaeological domain. The framework is structured in three important parts in the knowledge generation process: data, process and interaction.
We have analyzed archaeological document collections from Incipit and other
open-data repositories[
Following this method, each pair of clauses is tagged by one coherence relation. After, the elements of the relation chosen are characterized in the clause analyzed. Finally, it is necessary to validate the elections chosen in each analysis. In our study the original author of each document form Incipit repositories was contacted and asked to validate the outcome of the analysis
The analysis results indicate that there are types of coherence relations more
common in the analyzed texts than others, namely relations based on combining values of
several attributes to build the argumentation. This type of inferences is related to
combinatory tasks, and some data-mining solutions such as rule-association
algorithms[
This analysis allow us to characterize in an analytical classification the argumentation methods detected in the study, based on the objective of the user and the type of inference that was used.
The classification obtained as a result of the analysis is explained in the Figure 2, with the corresponding mappings with Hobbs relations. It is used as basis to formalize the cognitive processes managed by the user in archaeology. 4.2
Two test-sets has been designed and implemented to extract evidence of the assisting possibilities through software in the archaeological domain.
The first test contains multiple choice questions about reasoning modes in archaeology based on real data, tasks related to knowledge generation and initial reasoning used to process data in large datasets (See Fig.3). There are plans to publish all results and conclusions from this test-set.
The second test-bed includes multiple-choice questions about ten visualizations of
real archaeological data sets implemented through well-known visualizations
techniques (bars, charts, tree-maps[
The proposed framework will be developed as an integrated metamodel covering the three areas defined above: data, process, and interaction.
Metamodeling the archaeological cognitive processes explained in this paper and
extracting applicable interaction primitives are our next goals. In the first case,
situational software engineering[
In the second case, the empirical studies being carried out indicate the necessity of different levels of interaction and visualizations, depending on the inference type involved. We continue to work in this direction.
In addition to the integrated metamodel, the implementation of a software prototype is planned, with the objective of testing the solution and systematically analyzing the improvements achieved in the knowledge generation process by assessing the detected problems, their impact and the obstacles encountered.