The Intelligence That Reads Citizen Complaints and Routes Them with Meaning
A metropolitan municipality receives thousands of complaints and requests every day. Coming from call centers, emails, mobile apps, and the field, these submissions are scattered, free-text, and often disconnected from one another. Having an operator read each one and forward it to the correct department with the right context creates a massive burden. Metropolitan Municipality is a public complaint and request intelligence platform we developed on the Phexum Analytic Platform to eliminate this burden. It gathers submissions from different channels into a single smart system, reveals the meaning and recurring problems inside them using NLP, and sends not raw data but contextualized analysis to the relevant department.
Featured Results
The system automatically normalizes scattered submissions from call centers, emails, mobile apps, and field entries, consolidating them in a single center. Every submission transforms into a standard and analyzable data structure. The NLP layer automatically extracts entities from the text, such as addresses, neighborhoods, license plates, institutions, and people, routing the submission to the correct department and enriching it with root cause analysis. Anomaly detection catches similar complaints surging in the same region in a short time, highlighting critical situations. Users can query the data using natural language without needing technical knowledge. When transferring to CRM, forms are automatically filled by NLP, maximizing operators' work speed.
Who the Client Is and the World They Live In
We developed this platform for a metropolitan municipality that manages thousands of citizen submissions daily. The main difficulty of complaint management in the public sector is the scattered nature and free-text format of submissions. A citizen writes, "There is a water cut in [Neighborhood Name] this morning." Hidden within this sentence is a location, a problem type, and time information. Legacy systems, however, see this purely as a block of text. An operator manually reads and categorizes every submission, assigns it to the right unit, and enters it into a form. This process is both slow and misses the most important signals due to its isolated perspective. Ten complaints that seem ordinary when viewed individually might actually be the harbinger of a problem bursting in the same neighborhood. Metropolitan Municipality solves both these problems at once: contextualizing the submission and making patterns visible.
Cross-Cutting Concepts and Architecture
The first issue running through this entire platform is transforming text into meaning. To a machine, a complaint arrives merely as a string of characters; the system's job is to turn it into structured meaning. The technical term for this is Named Entity Recognition (NER). This method automatically recognizes and tags addresses, neighborhoods, license plates, institution names, and people from free text. From the sentence "The vehicle with license plate 34 ABC 123 constantly parks in the same spot," Metropolitan Municipality extracts the license plate, the recurring behavior, and the location relationship. The benefit of this extraction is twofold. On the one hand, the submission goes to the correct department with its root cause context. On the other, the extracted entities automatically fill CRM forms, saving the operator from manual data entry.
The second issue is moving from a single event to a pattern. A single complaint is an isolated event; complaints over time form a pattern. The system analyzes submissions not just as individual records, but as patterns on a time axis. It marks situations like a rapid increase in similar complaints regarding the same neighborhood as an anomaly and generates a high-priority alert. This approach is data-driven, dynamic prioritization that goes beyond the classic fixed "priority level" logic. On top of this, we added a natural language querying (free text analytics) layer that requires no technical knowledge. A question like "On which days and in which neighborhoods did water cut complaints increase this month?" is answered as a table, a time chart, or a map. The real benefit of this layer is that it takes data access out of the hands of the technical team and places it directly into the hands of the decision-maker.
How the Story Was Built
We built the system not as a separate infrastructure from scratch, but on top of the Phexum Analytic Platform. This decision was decisive from the start. Thanks to the platform's data processing and decision support capabilities, the collection, contextualization, and analysis of submissions could work as a single unified whole. Our goal was not merely to collect submissions, but to accelerate decision-making processes by extracting the meaning, recurring problems, and critical signals within them. Therefore, we designed the system not as a record repository, but as a living intelligence layer that generates context from every incoming submission.
Building Blocks of the Solution
The multi-channel consolidation layer normalizes submissions from different sources and transforms them into a standard data structure in a single center. The NLP and entity extraction layer automatically detects and structures entities in the text like locations, license plates, institutions, and people. The smart routing and root cause layer analyzes the background of the submission while assigning it to the right unit and enriches it with context. The anomaly detection and triage layer catches patterns over time and highlights critical regions. The natural language querying layer makes the data queryable as tables, charts, and maps. The timeline and trend analysis visualizes which problems are increasing on dashboards. The CRM integration layer automatically catches entities and fills forms, minimizing data entry.
A single complaint is an isolated event — complaints over time form a pattern.
The Impact We Created
Metropolitan Municipality transformed scattered and free-text citizen submissions into an intelligence layer that reaches the right department contextualized and can be queried by the decision-maker using natural language. Operators achieved maximum speed by being freed from manual reading and form filling. Critical regions were noticed early through pattern analysis. The institution turned a pile of complaints from a mere burden to be managed into a data source to understand the city.