Talk To Data

A web application that lets you analyse your data simply by asking questions in plain language, no coding required.

Context

Talk To Data

Talk To Data is a web application that transforms the way organisations access the information held in their data. Instead of requiring technical expertise, SQL queries, or specialist analysts, users upload an Excel file, a CSV, or connect to a database and ask questions in plain language as if talking to a colleague: "Which products grew the most this quarter?", "What is the turnover rate by department?", "Where are our biggest delivery delays?". The system interprets the question, identifies the relevant data, runs the analysis, and returns a text answer accompanied by the most appropriate chart for the data type, all within seconds and in real time. The result is a democratisation of data access: anyone in the organisation, regardless of their technical role, can get answers from their datasets without any intermediary.

What we did on it...

Vision

Project conception

The project grew from the observation that most company data remains inaccessible to most people. Databases contain strategic information, but querying them requires skills that few people have. Traditional reports are static and cannot answer specific questions. The goal was to build a tool that removed this barrier entirely, making data analysis available to anyone capable of asking a question.

Structure

System architecture

The system consists of three main components working independently and in coordination: a browser-based web interface, an AI engine that interprets natural language questions and generates analyses, and a database layer for temporary management of uploaded data. The response process runs in four sequential steps: question comprehension, relevant data identification, analysis execution, and optimal visualisation generation. Each component is designed to be independently replaceable, ensuring long-term evolutionary flexibility.

Interface

Application development

The interface is built with real-time streaming that keeps the user updated step by step during processing. Results are presented with explanatory text and automatically generated charts: bars for comparing categories, lines for time trends, pie for percentage distributions, scatter for correlations between variables. The system autonomously selects the most appropriate chart type for the question and data, without the user needing to specify it. Uploaded files are automatically deleted after 4 hours, ensuring data privacy and security.

Launch

Testing and release

The system was tested with real datasets from diverse business contexts, checking answer accuracy, relevance of generated visualisations, and processing times. The release is available both as a demo with sample datasets and with direct upload of real company files, enabling an immediate evaluation in the client's own environment.