T-LAB Plus 2020

T-LAB Plus 2019
6 February 2019

T-LAB Plus 2020

T-LAB Plus 2020

was released on October 3th 2019

Here is a short list of the most significant improvements made in this version of the software.

1 - The main menu, which has been redesigned, is now easier to use (see picture below).


2 - All T-LAB tools which include the 3d bubble chart option (e.g. Cluster Analysis, Correspondence Analysis, SVD, etc.) now use a new type of visualization which is highly customizable (see pictures below).


3 - In the Co-Word Analysis tool – as well as in other tools which produce word embeddings (e.g. SVD) - the visualization technique called t-SNE (t-Distributed Stochastic Neighbor Embedding) is now available which separates groups of associated words efficiently (see picture below).


4 - A new tool named Texts and Discourses as Dynamic Systems has been added which integrates various state of the art algorithms already present in T-LAB with the Recurrence Quantification Analysis (RQA).


In particular this new tool allows us to verify how texts are organized in time, how the recurring themes and the sequential order of utterances relate to each other and how similarities and differences between them evolve in time. For these reasons this tool – more than other T-LAB tools - challenges the divide between qualitative and quantitative approaches in text analysis.

Below are some pictures concerning the new features, each of them preceded by a short description marked with an asterisk.

(*) 3D real time waterfall chart where every text segment is described by as many histograms as the number of the available topics/themes.


(*) Recurrence plot where each dot refers to the similarity between a pair of text segments.


(*) Recurrence plot concerning the time series of topics/themes.


(*) Real time chart (heat map series) where each dot refers to the probability assigned to the relationship between a text segment on the abscissa and a topic/theme on the ordinate.


(*) Data table with similarity measures concerning all pairs of corpus subsets.


5 - Among the algorithms for cluster analysis, there is now also a version of Hdbscan available (see picture below)


6 - When importing a dataset – either in tabular format or not – consisting of lots of texts the user is now also allowed to tag them by means of a Unique Identifier, i.e. an alphanumerical string which will be added to the analysis outputs. Moreover, through the import/export options, any unique identifier list can be modified at any moment (see pictures below).
N.B.: The list of unique identifiers may also consists of proper names, geographical names, names of books etc., where each name must be a unique string – without blank spaces – up to 50 characters long.


7 - When using the document clustering (i.e. Thematic Document Classification) tool a new option is now available which allows the user to easily check the similarities (i.e. Cosine measures) between pairs of documents and easily export the corresponding data table (see picture below)