Approaches
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
136 137 138 139 140 141 142 143 144 145 146
In texts, the links between words (concepts) are implicit. Hence, extracting a network of
concepts from a text, and classifying this network via the meta-matrix ontology, requires
an inference process. The links, or relations, between concepts must be extracted based
on the semantic, syntactic, and contextual information given in a text (Carley, 1986;
Carley, 1988; Popping, 2003). Making the meta-matrix approach available for NTA can
provide analysts with a novel technique for extracting textual networks that reveal the
relationships within and between the elements that compose a network and that were
classified a priori according to the meta-matrix model. The features of the textual data
that are relevant to the analyst can then be represented as a network structure of the meta-
Table 2: Original meta-matrix conceptualization
Based on Carley (2002, 2003)
Meta-Matrix entities People Knowledge/
Resources
Events/
Tasks
Groups/
Organizations
People Social network Knowledge
Network/
Resource
Network
Attendance
Network/
Assignment
Network
Membership
network
Knowledge/Resources Information
Network/
Substitution
Network
Needs
network
Organizational
capability
Events/Tasks Temporal
Ordering/
Task Flow/
Precedence
Institutional
support or
attack
Organizations Interorganizational
network
matrix entity classes and the connections between these classes. Such a network makes
the structure of social systems, which is implicitly contained in texts visible and
analyzable.
How did we combine and formalize the meta-matrix approach and map analysis technique,
which is a specific type of NTA? We utilized the meta-matrix model as an extension of
NTA in general and map analysis in specific by instantiating the following five step
procedure:
1. Concept Identification: identify the concepts in texts that are relevant to the
analyst’s research question. As part of this process, the analyst may first want to
generalize many text-level concepts into higher-level concepts.
2. Entity Identification: define an ontology for capturing the overall structure
described in the text. We use the basic meta-matrix. Other analysts may wish to
adapt this to their research question. Note, step 1 and 2 can also be done in reverse
order.
3. Concept Classification: classify the identified concepts into the relevant entity
classes in the meta-matrix. Given the vagaries of the language it may be that some
concepts need to be cross-classified in two or more entity classes.
4. Perform Map Analysis: automatically extracting the identified concepts and the
relations among them from the specified texts. This results in a map or conceptual
network. Since the concepts are classified by entity classes, the resulting concept
network is hierarchically embedded in the ontology provided by the meta-matrix.
In essence then, there are three networks. First, there is the concept network where
the nodes are concepts (many of which are higher-level concepts). Second, there
is the entity network where the nodes are the entity classes and the links are the
connections among and between the entity classes. Third, there is the network
embodied in the meta-matrix thesaurus, connecting concepts in entity classes to
concepts in the same or other entity classes. Finally, there is the network (embodied
in the meta-matrix thesaurus), connecting concepts to entity classes.
5. Graph and Analyze Data: the final step is to take the extracted data for each text,
the network, and graph and analyze it in general and by cells in the meta-matrix. As
part of this analysis, the resultant networks from different texts can be combined
and compared. Note the analysis can occur at the concept network level (map
analysis), the entire meta-matrix level (meta-matrix text analysis), and the sub-cell
level (sub-matrix text analysis).
We refer to these five steps as the method of meta-matrix text analysis. With this novel
technique we hope to contribute towards the analysis of complex, large-scale data and
social systems and providing profound multi-level access to the meaning of textual data.
We note that these steps begin to bridge the gap between NTA and a more interpretive
analysis of texts. The meaning of concepts is revealed by virtue of other concepts they
are connected to. In the meta-matrix approach, the meaning of concepts is revealed both
by what other concepts they are connected to and by what type of entity classes into
which they fall.