When it comes to applying artificial intelligence (AI) to the translation field, a popular misconception is that AI only has significance for linguistics. (i.e. using computer algorithms to parse the source language text into phrases and sentences and then convert these segments into another language.) While AI plays an important role in today’s MT technologies, especially the latest neural models from Google, Amazon, and Microsoft, it also has real and significant applications in many other aspects of the language translation and localization process. These applications include AI powered file analysis, estimating turnaround times, language resource management, translation provisioning, translation reuse, linguistic review, and risk management.
Interested in learning how AI can make your enterprise translation operations better and smarter? Please contact firstname.lastname@example.org for a live technology demo.
AI Powered File Analysis
File analysis is an integral part of the localization process. It extracts text into a plain text format so linguists can easily translate on desktop and mobile. The analysis then calculates the word count for workload and cost estimates, leverages legacy translations for linguistic consistency and cost savings, and identifies terminology for translation technical accuracy. Traditionally, file analysis has been performed by human operators using CAT tools like Trados to OmegaT to process various file formats such as Word, PDF, PowerPoint, Adobe InDesign, XML and others to get the word counts for new words and fuzzy matches. Next, they manually apply the rates to produce the translation quote in an Excel or PDF format.
However, this manual process involves multiple file transfers leading to possible misplacement and calculation errors. The process is also time consuming, which is why traditional localization companies usually have a hard time providing translation estimates instantly, often requiring a day or more to provide even the simplest translation quotes.
Stepes AI powered translation management system automates the file analysis tasks by automatically detecting the source text language, leveraging TMs on the cloud, and providing translation quotes in real time. In addition to regular file types, Stepes has expanded our AI capabilities to include scanned documents and image files which extract text in real time, on-demand. This highly automated process is a game changer because Stepes customers no longer have to wait for 24 hours or longer just to receive a translation estimate. In today’s digital economy, a single day’s delay can cause lost business opportunities with millions of dollars in lost revenue.
AI Powered Translation Provisioning
Believe it or not, translation provisioning, or assigning translators to actual projects, is a big part of the professional translation process. This is the task that a translation company’s project managers perform in order to send files needing translation (or translation kits) to individual translators. Since most of the localization industry today works with freelance translators who are based in-country, email messages and FTP (or portal) transfers have been the norm to hand off files to linguists. However, this process has several important drawbacks:
- There is no guarantee translators are notified right away, causing unnecessary delays which can be critical for urgent translation projects. In fact, it’s quite common for a project manager to send the same files to multiple translators and hope someone will respond quickly.
- This process is highly manual and time consuming especially if the number of languages involved is large. The manual emailing process is also prone to human errors causing incorrect files to be sent to the translators resulting in lost time and quality.
- Project managers working offline often don’t have full visibility of a translator’s availability and/or subject matter expertise. This means the translation job is sent to a linguist who is already busy with something else or not the most qualified to take on the translation assignment, leading to missed deadlines and poor-quality results.
Achieving Translation Scale using AI
With Stepes, enterprise customers can decide how fast they need content translated regardless of the amount of words involved. For example, it’s typical for a translation company to translate 3,000-10,000 words a day so a project with 300,000 words would require 4-5 weeks or longer. However, sometimes, companies need this content translated quickly with quality in less than a week in order to meet urgent business deadlines. This is where conventional translation workflows tend to struggle leading to missed delivery deadlines or poor quality, or both.
Stepes AI enabled project management workflow solves this bottleneck by using its distributed project assignment APIs that performs the following:
- Splits the extracted segments into multiple jobs to be assigned to different translators. The actual number of jobs depends on the total word count and client specific deadline.
- Identifies available linguists on its large network of translators based on subject matter expertise and performance ratings. Job notifications are pushed to the linguists automatically on their mobile and through email.
- Performs Dynamic Terminology Management (DTM) during actual translation to ensure linguistic consistency and technical accuracy.
- Automatically merges translated segments to create the final document.
With Stepes AI translation solutions, customers have the ability to set their translation project deadlines easily such as one hour, 24 hours, or 3 days without having to worry about language quality and potentially missed deadlines.
Intelligent Language Resource Management
One of the most important benefits of Stepes’ AI powered translation system is that it uses the modern network to most efficiently organize and manage linguistic resources on a large scale. Modern translation businesses depend heavily on quality linguistic talents. However, evaluating, qualifying, and managing these translators is a labor-intensive task that’s highly prone to errors, leading to poor translation results.
Stepes’ AI powered translator management capabilities automate many of the manual processes associated with language resource management for results that are better than ever. Specifically, our intelligent screening process automatically checks translators’ IP addresses, phone numbers, and translation experiences to ensure the highest quality and reliability while reducing risk. Stepes uses sophisticated algorithms to classify linguists based on subject matter expertise, experience, and translation performance ratings. These factors are used when the system assigns translation projects to each translator for the best linguistic results and on-time performance.