T-LAB Plus 202016 September 2019
T-LAB Plus 202216 October 2021
T-LAB Plus 2021
was released on October 14 2020
Here is a short list of the most significant improvements made in this version of the software.
1 -Now the main analysis tools include a quick access gallery of pictures which works as an additional menu and allows one to switch between various outputs with a single click (see pictures below). This way, while the classical menu on the bottom left – which offers more options - remains the main reference, non-expert users can easily familiarize themselves with the software.
2 - More supervised learning options are available in several tools for thematic analysis and now the user can easily obtain outputs like confusion matrices and precision/recall metrics. In particular: (a) the Dictionary-Based Classification tool allows one to test any model on labeled data (e.g. data which includes themes obtained from a previous qualitative analysis) with just a few clicks; (b) any cluster partition obtained through the T-LAB thematic analysis tools (e.g. Thematic Document Classification) can now be tested with a Naïve Bayes classifier (see pictures below).
2-a: Dictionary-Based Classification
2-b: Thematic Document Classification
3 - A new wizard has been added which allows one to easily classify new documents according to a pre-existing model (i.e. any categorical variable) and also to compare any new document with all documents included in a corpus already analysed (see pictures below).The possible uses of this new tool include sentiment analysis and plagiarism detection.
N.B.: When the user wishes to classify a dataset of new documents by using a supervised method, the dataset must be imported by T-LAB and then analysed by using a previously generated dictionary. To this purpose, the Thematic Document Classification can be used, both for generating a dictionary of categories (i.e. unsupervised method) and for performing a supervised classification.
4 - When importing new texts, selecting the language of the corpus is now easier. Moreover, the pre-processing setup includes two more options for social media data: (a) automatically remove hyperlinks from texts and (b) consider each text line as a single co-occurrence context (see picture below).
5 - The Co-Word Analysis tool has been further improved. In fact, now it allows the plotting of customizable charts with up to 500 points (i.e. key-words), and a 3D PCA (Principal Component Analysis) is also available (see pictures below).
N.B.: When too many points are displayed, by using the left click of the mouse, the user can zoom in on any section of the chart (see pictures below).
6 - The Specificity Analysis tool now offers more options, which are all accessible through simple binary choices. In particular, any comparison between corpus subsets (which can also be text documents) can be performed both by using the occurrences of ‘lemmas’ and the occurrences of ‘words’. Moreover the user is now enabled to easily evaluate similarities (i.e. Cosine) and differences (i.e. Inter-Textual Distance) between corpus subsets (from 2 to 150), and so to detect duplicate and near-duplicate documents (i.e. evidences that may also indicate plagiarism; see pictures below).
N.B.: The above picture shows that, by comparing the lexical profiles of candidates in the recent US debates, the highest similarity (Cosine: 0.671) is between Kamala Harris and Mike Pence. And - interestingly - the lowest similarity (Cosine: 0.044) is between Donald J. Trump and Mike Pence.
7 - When using co-occurrence analysis tools like Word Association and Sequence Analysis, the user is now allowed to test the significance of all links between key-words by means of the Test-Value (Lebart L, Morineau A., Piron M. - 1995).
N.B.: In the 2021 version the Test-Value can also be used to verify whether a statistically significant relationship exists between key-words and topics (or themes).
8 - When using the Comparison between Word Pairs tool, a new chart is now available which allows the user to compare the various association indexes and export their values.
10 - The Export custom tables tool (old name: ‘Contingency tables’) now allows the user to visualize and export both occurrence and co-occurrence tables with up to 5,000 columns. This way, by using other software, expert users can easy import and analyse the data tables provided by T-LAB.