Specificity Analysis

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. In particular, starting from the 2021 version, a quick access gallery of pictures which works as an additional menu allows one to switch between various outputs with a single click. Moreover the user is enabled to easily evaluate similarities (i.e. Cosine) and differences (i.e. Inter-Textual Distance) between corpus subsets (from 2 to 150), and so also to detect duplicate and near-duplicate documents (see pictures below).

This T-LAB tool enables us to check which lexical units (words, lemmas or categories) are typical or exclusive in a text or a corpus subset defined by a categorical variable, as well as to check the 'typical contexts' of each analysed subset (e.g. the 'typical' sentences used by any specific political leader).

In detail:

The typical lexical units, defined for over-using or under-using, are detected by means of the chi-square or the test value computation.

The typical elementary contexts are detected by computing and summing the normalized TF-IDF values assigned to the words which each sentence or paragraph consists of.

Specificity Analysis allows us to carry out two types of comparisons:

1- between a part (e.g. the subset "A") and the whole (e.g. the corpus under analysis, "B");

2- between pairs of corpus subsets ("A" and "B").

In either instance Specificities involving both the intersection (tipical words) and the differences (exclusive words) can be analysed.

The computation modalities are shown in the corresponding glossary entry.

The considered lexical units can be all (automatic settings) or only those selected by the user (customized settings).

The four types of possible comparisons are as follows:

1.1 - part/whole: "typical" lexical units

Table reading keys are as follows:
- LEMMA = specific lexical units (over/under used);
- SUB = occurrences of each LEMMA in the subset;
- TOT = occurrences of each LEMMA in the corpus or in the two compared subsets (see 2.1 below);
- CHI2 = CHI square value (or VTEST = Test Value);
- (p) = probability associated with the chi square value (def=1).

By clicking on the listed items it is possible to create various charts (see below).

1.2 - part/whole: "exclusive" lexical units

2.1 - subset/subset: "typical" lexical units

2.2- subset/subset: "exclusive" lexical units

For each targeted subset it is also possible to check its 'typical' elementary contexts, the specificity of which is a result of the computation of normalized TF-IDF values. More specifically, the 'score' assigned to each elementary context (see the picture below) results from the sum of TF-IDF values assigned to the words which it consists of.

All contingency tables can be easily exported and allow us to create various charts.
Moreover, by clicking on specific cells of the table (see below), it is possible to create a HTML file including all elementary contexts where the word in row is present in the corresponding subset.

Eventually, by clicking the appropriate button (see below), a dictionary file with the .dictio extension is created which is ready to be imported by any T-LAB tool for thematic analysis. Such a dictionary includes all typical words of the selected categorical variable.