How to Build a Private ChatGPT Using Open-Source Technology? Download our free white paper.

Uside: Analyze employee surveys to monitor conduct of change


Ready to go for the gold with Lettria?

Analyze employee surveys to monitor conduct of change

Uside is the first consulting firm specialized in cultural and behavioral changes.

With its unique expertise in behavior based on a scientific methodology, Uside's team is made up of over 70 experienced senior consultants working with two-thirds of the CAC 40 companies. Uside creates and develops the conditions for a positive dynamic of adaptability by making people the driving force behind sustainable performance.

Lettria, in partnership with Uside, worked on a solution to automate employee survey analysis for a company in the energy sector.

Analyze emotions and identify solutions from a large panel of verbatims

Faced with a massive reorganization of its teams, our client wanted to interview its managers and senior managers in order to collect their feelings and to consider solutions to improve internal change management.

More than 200 workshops were organized during which 1,800 participants shared their opinions on "good practices" to be implemented. These workshops were divided into 3 parts: first, 3 good practices were submitted, followed by a collective brainstorming and completed with more free suggestions.

The workshops were a great success, yielding over 1,300 verbatims. Manually processing so many texts is not a tenable solution; the additional processing costs and the risk of human error are too great.

Lettria and Uside therefore proposed a solution to automate the analysis and synthesis of verbatims.

Automating the analysis and summarization of verbatims using NLP

To guarantee relevant data processing, Lettria proposed an analysis with 3 technical approaches to determine the main axes, or trends that emerge from these workshop feedback sessions.

  • Facilitating reading thanks to lemmatization

Thanks to lemmatization, i.e. the morphological analysis of a text by simplifying words, we were able to standardize the text. The objective was to gather synonyms and close formulas to build a more generic and harmonized word cloud. With this perspective, it is possible to identify the key elements of the verbatims and to save time without getting lost in the text.

  • Isolating key words using Pos-Tagging

To refine this process, Lettria uses Pos-Tagging to retain only verbs and common nouns. Conjunctions, determiners and other minor tags are not processed. Only meaningful words are kept and analyzed.

  • Contextualize with the dependency links between words (Dependency Matcher)

Once the text is segmented and the main words are isolated, we use the Dependency Matcher to give meaning to the different components of the text. Adjectives are linked to the corresponding nouns, object complements to verbs, etc., in order to contextualize the main ideas. Without Lettria's solution, it is impossible to combine these words and accelerate data processing. It would therefore not have been possible to (quickly) identify the essential axes.

  • Analyze emotions

The objective is to understand, through sentence constructions, adjectives, etc., what feelings are associated with each issue evoked in the surveys. In this way, it is possible to determine the positive or negative opinions of employees based on the topics and comments.

In short, the most present and important words stand out and are associated with a color that represents a feeling. This is how the main analysis axes are determined.

Best practices white paper

The project took place over 3 weeks and required 1.5 months of development. The processing of the verbatims allowed the company to produce a white paper of "best practices" to facilitate its reorganization and the acceptance of the change by its employees.

The tool offered as a turnkey solution, using Lettria technology in part, also makes it possible to update, refine, modify or add new verbatims to perfect the analysis. New verbatims are added directly to the interface. The conclusions obtained can thus be enriched and adjusted as the project progresses according to management feedback.

Build your NLP pipeline for free
Get started ->