Co-Word Analysis and Concept Mapping

N.B.: The pictures shown in this section have been obtained by using a previous version of T-LAB. These pictures look slightly different in T-LAB 10. Also: a) when the 'automatic selection of key terms' is selected, different colours are used for different groups of items in the MDS map; b) the visualization technique called t-SNE (t-Distributed Stochastic Neighbor Embedding) has been added; c) there is a new button (Graph Maker) which allows the user to create several dynamic charts in HTML format; d) by right clicking on the keyword tables, additional options become available; e) a quick access gallery of pictures which works as an additional menu allows one to switch between various outputs with a single click.
Some of these new features are highlighted in the below image.

This T-LAB tool allows us to find and map two kinds of relationships concerning word co-occurrences:

A - between single key-words (lemmas or categories), if their number does not exceed 500 elements (min 10);
B - between/within little clusters (i.e. Thematic Nuclei), if the number of key-words selected exceeds 100 elements (max 3,000).

The user may choose which association index to be used and, for option B only, he may also choose both the maximum number of clusters to be obtained (from 50 to 100) and the maximum number of key-terms within each cluster.

The computation process includes the following steps:

1- building a co-occurrence matrix (word x word);
2- computing the selected association indexes (Cosine, Dice, Jaccard, Equivalence, Inclusion, Mutual Information);
3- hierarchical clustering of the dissimilarity matrix;
4- building a second dissimilarity matrix (cluster x cluster);
5- graphic representation by multidimensional scaling and correspondence analysis.

N.B.:
- in "A" cases ((see the below image), the user can review the key-term selection and T-LAB doesn't carry out steps 3 an 4;

- the quality of results depends on a thorough selection of key-words;
- as the multiwords unclassified by T-LAB are specific cases of co-occurrence and the "B" option treats them like little clusters (e.g. "Twin" + "Towers"), the user is advised to resolve these cases during the pre-processing phase. Anyway, without repeating the corpus importation, it is possible to make changes by means of the Dictionary Building function (e.g. by assigning the label "Twin_Towers" to the two different items "Twin" and "Towers");
- by clicking on the appropriate buttons all data tables can be checked (see the picture below).

When the automatic analysis is over, four kinds of charts are available (see below) and each of them can be customized by using the appropriate dialog box (just right click on the chart).

1 - MDS Map

2 - Factorial Analysis of Correspondences

3 - Association Diagram

4 - Diagram of Centrality-Density measures (after a cluster analysis only)

In particular, the results obtained by Correspondence Analysis can be mapped using the coordinates of the first ten axes (see "A" below).
As T-LAB allows us to verify the Test Values of each factor (see "B" below), this kind of output can be useful for an accurate interpretation of the relationships between cluster and/or key-words.

The charts can be explored and customized in the following ways:

A further option allows us to select the items (i.e. the 'nodes') for Network Analysis (see the image below, step 1 and 2), to export the corresponding adjacency matrix (step 3), select the links on the basis of their range of probability value (step 4) and export different types of files (step 5) which can be edited by software such as Gephi, Pajek, Ucinet, yEd and others.

N.B.: In T-LAB 10 the following window has been replaced by the Graph Maker tool.

There are available three tables which can be exported by this T-LAB tool:

1 - "Cluster Membership" table (see below) deals with the hierarchical aggregation of words within each cluster;

2 - "Summary" table (see below) includes the following measures:

- ECQ = Quantity of Elementary Contexts in which two or more word clusters are co-occurring;
- Centrality = average of association indexes concerning cluster relationships;
- Density = average of word association indexes within each cluster.

3 - "Association Indexes" table (see below) includes similarity measures of the between and the within cluster relationships.

Between	Within

N.B.:
- when a Cluster Analysis has not been carried out, the "Cluster Membership" table is not available, consequently the "Summary" is simplified and the "Association Indexes" table refers to word co-occurrences only;
- when exiting from this analysis, the dictionary of Thematic Nuclei (i.e. the list of labels assigned to each word cluster) can be exported and, after a thorough revision, can be imported by means of the Dictionary Building function. In this way the user will be able to perform certain second order analyses (i.e. analysis concerning "themes" or "concepts").