Semantic alalysis

How does semantic analysis of user-generated content look?

Semantic analysis implies the search for semantic connections and parameters in chat messages. When analysing messages, you can find insight points crucial for deeply understanding user behaviour, preferences, and attitudes.

Average semantic analysis includes searching for main topics that users discuss, grouping the messages by topics, and counting their shares.

Why do you need it?

Due to semantic analysis, you learn what your users are interested in, their requests and needs the platform and in-app chats can meet, and users’ everyday demands.

Semantic analysis of UG content can substitute direct interviews or focus groups with users because the number of respondents increases to the number of chat users. Their answers become more honest as they express their opinions intentionally, not answering the platform’s representative. If required, analysis can be simplified to search for messages' emotional undertones (Х% messages—positive, negative, neutral).

You can analyse all users’ messages for the period, messages of separate user groups (e.g., users with VIP status), or messages to separate events or titles.

You can read more about semantic analysis and how it differs from sentimental analysis on our blog: https://watchers.io/post/semantic-and-sentiment-analysis-what-are-the-best-sources-of-consumers-thoughts

How to produce semantic analysis through the back office

You can find the AI tool for semantic analysis on the admin panel: https://chatadmin.watchers.io/semantic

Click "New analysis" in the top right corner to start the process. The previous results are on the same page; you can always get back to them if needed.

Add your OpenAI key if you have one in Settings, or use the default one.

Then, add the prompt. You can use one of the saved prompts or add your own one, but be consistent and transparent in your task. Remember that the concept of this tool is distinguishing messages into groups. So, your prompt must include categories for grouping messages and explanations (ideally, with some examples) about they mean.

Then, you need to add messages for analysis. If you want to analyse all messages for the period, select "Choose a period," and all messages sent to chat during the chosen time will be analysed. If you need to analyse and categorise specific messages, you can upload your own CSV file, but follow the instructions of the admin panel regarding how to form this file.

Then click Start. You can download the table with the results when the analysis is ready. Except this, you will find the most popular words used by your users on the same page.