When Apple Maps first debuted, the company was heavily criticized for inaccuracies and errors. For example, New York’s iconic Madison Square Garden was categorized as a “nature area” simply because it contained the word “garden”. The application was so negatively received that Tim Cook even posted a public apology letter promising to improve the service.
Since then, Apple has made significant improvements to their software by correcting errors, building new features and even planning a major overhaul of the platform. This goes to show that in the navigation software war, the app with the most accurate data wins out.
Unbeknownst to the general public, human-annotated location data is what drives the success of the best navigation software. Used to enhance and verify location data, a human touch provides tremendous value for end users. Listed below are three ways human-annotated data improve navigation software.
Local Search Relevance
The core function of search engines is to provide relevant and accurate results. In the case of navigation apps, local search is aimed at finding something within a specific geographic area (e.g. “Chinese food in Tokyo”). Making local search results stand out from others can be a huge advantage for navigation applications. Verifying the accuracy of local search results is of critical importance to retaining users.
Search relevance refers to how well fetched results match what a user is looking for. Human knowledge is the single most important resource for improving local search relevance.
For a search algorithm to achieve high precision and high recall, it needs to be trained consistently using high-quality human annotated data. To create that data, human evaluators work behind the scenes to rate millions of search queries alongside their results.
Another important factor for navigation app success is the accuracy of business information. Having complete, detailed, and up-to-date business information also contributes to local search relevance. Likewise, failure to keep accurate local listings can result in users losing confidence in the search engine’s data and turning to its competition for future queries.
The best way to achieve data accuracy is to have in-market experts verify metadata and geographic placement for each location. Humans review, verify, clean, and label all types of data to ensure business listings were as accurate and relevant as possible for its customers in all markets.
Another essential functionality of map & navigation applications is the ability to quickly find the most efficient driving, cycling, and walking directions. Particularly for underdeveloped markets, precise localization for real-world scenarios is a challenge for mapping applications.
Humans have a large role in this process. In-market software testers can help ensure that navigation from point-to-point is accurate, safe, and timely. Local evaluators are staffed so that companies can access knowledge only a local expert can have.
Despite significant advances in map and navigation software, tech’s biggest players still rely on humans in the loop to stay ahead of the competition. Interested in collecting map data for machine learning? Lionbridge AI can provide geo-local AI training data in over 300 languages. We have over 500,000 qualified professionals working on our platform in all major time zones and almost every country, so we can keep up to speed with your data needs.