loader

Case Study: Research

The text mining techniques allows to capture and understand people thoughts, opinions and preferences on social networks and web with respect to a certain discussed topic.

In particular, a vertical analysis focuses on a specific market sector of interest, on a certain brand and its competitors, or on desired products of that industry. All the tools are provided to understand the business deeper from a web perspective.

Project Details

Sector: Research
Period of analysis: 1 April - 8 May
Total analysed mentions: 100.000
Sources: Twitter, Blogs, News, Forums

Topic Discovery

By means of the Topic Discovery analysis it is possible to understand the most discussed topics of some complex textual data. This allows also to capture the underlying justification to the sentiment and NPS evolutions over time, and to perform various types of cross-analysis.

The Wordcloud is usually a good starting point to discover the main trending topics people are talking about on the web. It shows hashtags, keywords or sentences with the size of each word indicating the volume of use relative to the other, and the color showing their “trendiness”.
In this research example, the two wordclouds show the main economic topic discussed by italian people during the period of Coronavirus.

The Topic Model, based on LDA (Latent Dirichlet Allocation) analysis, is another useful unsupervised method that helps to identify the interesting topics that underline the textual data.
The three different flower plots show three different topics italian people were talking about during the Coronavirus crisis.

Topic Discovery

The General View shows the estimated aggregate percentage of discussion for each topic that has been identified as being “trendy”.
The vast majority of posts discussed of MES (or ESM), while just a relative small portion of comments talked about Green Recovery and Lockdown.

The Daily View instead allows to see how each trending topic varies with time, revealing possible interesting trends.
It is usually of interest to analyse deeper the different topic paths using a qualitative peak-and-trough analysis. This specific analysis helps to reveal the highs and donws of each topic evolution.

Web Opinion Index Analysis

The Uncertainty Index and the Panic Index are the two indicators created by the T-Voice Coronavirus Observatory with the aim to estimate the levels of uncertainty and panic from people opinions during the coronavirus health and economic crisis.

Both are based on a 1 to 4 scale of panic or uncertainty rate respectively (where 1 means no panic or uncertainty and 4 implies high panic or uncertainty), which are then normalized to 0-100 through the min-max method.

The results are two indexes that can be directly used to compare different countries, regions or even cities, both at an aggregate level and over time.

In particular, values between 0-33 can be interpreted as acceptable levels of panic or uncertainty, while values above 33 are signals of worry.

Web Opinion Indexes: the new KPI

A Web Opinion Index (WOI) is a synthetic indicator that translates unstructured textual data into structured quantitative information that synthesizes the topic of interest. Essentially, web data are analysed and converted into numeric values by means of the artificial intelligence, and then normalized through well-known statistical techniques.

The result is an index that generates valuable business insights and is both easy to understand and to use. This new KPI, the WOI allows to monitor and compare the evolution of relevant topics.

WOI can be created for every desired sector, brand or products.

OUR ANALYSIS

Recent Analysis