Utilizing AI for multilingual Traffic Announcements – Possibilities and Limitations

Helsinki's ambition to offer better and more user-friendly traffic services is a timely and important goal. A solution utilizing traffic data from various sources to synthesize and deliver information audibly using Text-to-Speech (TTS) technology offers a new way to improve both accessibility and information dissemination for various user groups. This article examines the opportunities and challenges of such a solution, especially in Helsinki, where multilingualism and personalized services are key values.

Opportunities

Support for Multilingualism

Helsinki is a multicultural and multilingual city, with a significant portion of its population speaking languages other than Finnish, such as Swedish, English, Estonian, Somali, Russian, and others. At the beginning of 2022, 17% of Helsinki’s population were non-native speakers, a figure expected to grow to 215,702 by 2040. TTS technologies like Google Text-to-Speech, Amazon Polly, and Microsoft Azure TTS offer extensive language support, enabling traffic information to be read in multiple languages. This can assist tourists and new residents who have not yet fully learned Finnish or Swedish. Such a solution supports Helsinki’s goals to be an international and diverse city.

Customized Tone and Style

Modern TTS technologies allow the customization of tone and style to suit user needs. For example, traffic announcements in Helsinki can be tailored for different audiences: a formal tone for official messages or a more engaging style for children. Such flexibility enhances accessibility for diverse user groups, such as providing clearer, friendlier tones for families or concise, direct styles for commuters.

?Personalized Customizations

User profiles can enable tailoring traffic information to individual preferences. Users can prioritize specific routes or types of information, such as public transportation updates or cycling routes. In Helsinki, where transport modes are diverse, this feature can offer users relevant information for their daily needs.

Constraints

Quality of Translations

While multilingualism is a significant opportunity, automatic translations of traffic data can lead to errors, especially with context-dependent terminology or local expressions. Such errors are particularly problematic for critical traffic information like accidents or emergency arrangements.

Naturalness and Comprehensibility of Speech

Despite advancements, TTS technologies vary in quality across languages. Ensuring natural and comprehensible speech is vital in Helsinki, where trust in official communications is critical. Certain dialects or languages may sound unnatural, potentially affecting message reception.

Complexity and Resource Demands

Supporting multiple languages and customizations increases system complexity and resource requirements. Developing and maintaining such a system in Helsinki’s efficient public sector environment poses challenges, requiring significant investments in both technology and human expertise.

Real-time Data Processing

Traffic data changes frequently, and synthesizing multilingual real-time information may introduce delays. If TTS technology fails to deliver timely or accurate updates, users may miss critical traffic information. This is particularly crucial in Helsinki, where exceptional traffic situations impact daily life.

Traffic Data Usage and User Needs

According to Fintraffic’s 2024 large traffic survey, environmental and safety considerations significantly influence Finnish drivers. Most drivers aim to avoid unnecessary driving and adopt economical driving practices. However, improvements are needed in driver alertness levels. The survey also revealed that users value accurate, real-time, and centralized traffic information. Common sources include the HSL app, Google Maps, and radio. Challenges include fragmented information and update delays, highlighting the need for integrated solutions.

Example Implementation – AIWeiser Traffic Info

Overview

The AIWeiser Traffic Info solution reads traffic data from public APIs and generates traffic updates in multiple languages. These updates are converted to audio messages using text-to-speech generators. Developed in a Python-Flask-PostgreSQL environment, the application leverages Python’s extensive libraries for AI and machine learning development. You can access the solution here: https://aiweiser-8d03d07c06bd.herokuapp.com/trafficinfo

Data Reading and Message Formation

The application polls traffic updates from Digitraffic and stores them in a PostgreSQL database. Traffic messages are partly rule-based and partly AI-generated, ensuring accurate formatting and terminology in Finnish, Swedish, and English. Other languages are translated from English.

Voice Message Generation

Voice messages are generated using ElevenLabs and Google Translate TTS tools. ElevenLabs offers versatile voice options and allows voice customization. However, it is slower and costlier compared to Google’s gTTS library, which provides reasonable results for most languages.

?

?

Recommendations for Further Development

?

Enhancing Language Models

Investing in NLP models optimized for Helsinki’s multilingual environment can improve translation and pronunciation quality. Collaborating with language technology companies and universities can support such efforts.

User Testing and Feedback Collection

Continuous testing and user feedback are essential in Helsinki’s multilingual setting to enhance naturalness and user experience.

Managing User Profiles and Data Security

Developing secure interfaces for managing personal settings can provide tailored services while maintaining data privacy and security.

Summary

Developing an AI-assisted traffic information synthesis and TTS solution for Helsinki presents an excellent opportunity to enhance accessibility and usability for various user groups. However, addressing translation quality, naturalness, and system complexity is essential to meet the needs of residents and visitors effectively