Externship with TripleTen: Allcorrect Games

When looking for a job, you need a strong resume and a portfolio that will impress your potential employer. But where do you find projects if you’re still learning the ropes? TripleTen has the perfect answer – externships.

Externships are similar to internships, but they are part of the study program. They mimic the real-life tasks of companies or provide actual assignments. As a coding bootcamp student, you can join a team to create an actual website or mobile app. And here’s why you should definitely go for it.

Previously, we covered an externship at Arqlite, where a group of our software engineering students worked on a new landing page for the client. 

This time, however, our students took part in an externship at Allcorrect Games. Here’s their experience working with a real client! (Hint: They did very well!) 

Customer

Allcorrect Games is a Dublin-based game content studio with a team of native-speaking linguists and artists, that runs localization testing for game publishers. The studio’s clients are the gaming industry’s major players that like to reach out to new markets, and need the game to be translated and adapted to fit the cultural context. 

However, the studio’s role isn’t limited to adaptation. They also want to validate the idea with the game publisher and predict how many game players will benefit from the localization. 

“After submitting the deliverable - the translated content of the game - we used to lose track of how the localization impacted the actual end users because we work directly with publishers [of the games],” said Demid Tishin, Allcorrect Games founding partner. “So the task was to see how much help the translations were for the gamers. If we could accurately determine what problems end users have in different locations with game localization, then we can be proactive rather than reactive, and approach our clients with suggestions, for example, to add a specific language.”

Task

TripleTen partnered with Allcorrect Games to help them research the localization outcomes and potential. The team was tasked to identify user reviews and analyze them in order to draw valuable conclusions. Not all reviews were relevant: participants had to look into three top gaming platforms and find only those related to localization. 

“Mining [...] millions of reviews manually just to find these particular ones that mentioned the quality of localization is a nightmare,” admitted Demid. “We have always wanted to automate this completely, but the task was way above our grade.”

Process

To accomplish this, TripleTen gathered a team of eleven data science and data analytics bootcamp students. They had to draw a multistep algorithm to do the following:

• Extract and process reviews from Apple Store, Google Play, and Stream 
• Develop business logic for Allcorrect Games to reach their objective.

The workflow formed a basis for the roadmap, starting with an introductory webinar where the participants were introduced to the task and the best work practices. The coordinators then planned the milestones in the form of group round-ups to review the recent steps and discuss any blockers. There was also a pair programming webinar, which gave the team a sense of unity and productive work. Finally, after five weeks, the team was expected to develop a final presentation for the client. 

Here’s what the TripleTen students and the team leads agreed on:

1. Data preparation - Extract relevant data and arrange it into a dataset.
2. Create an algorithm to determine if the review is localized or not.
3. Build an algorithm for dividing localization reviews into positive and negative.
4. Create a dashboard for the business team.
5. Identify the structure and dynamics of rating-based localization reviews. In other words, find patterns for the algorithm to distinguish localization reviews from general ones.
6. Indicate the most common reasons for negative reviews and present them as a visual to the client.

Challenges

Before any project starts it is important to negotiate and agree on the expected outcomes. In the beginning, a client may be not sure of which outcomes will benefit their business most or have a very vague idea about it. It’s the project manager's job to ask questions and clarify future steps and help the client to find the best solutions. It will not only help the client to reach their goals faster but will also save data analysts or data scientists from ineffective work.

This externship highlighted the importance of that first step - communication and setting clear goals for the project. During the project, the client realized that they needed another type of outcome, and the emphasis changed from data analytics reports to data science analysis. This of course resulted in changing the workflows and plans.

Although most teams do their best to come to an agreement on the outcomes and presentation before the work has started, adjustments are unavoidable. During the externship, students got that experience first-hand. They were able to train their skills in rapidly changing situations, which undoubtedly contributed to their practical skills, knowledge of the pace of the tech business, and how to handle it all.

Outcomes

During the final presentation, TripleTen students were able to show their findings to representatives from Allcorrect Games. 

“The data analytics students highlighted a few languages that showed a growing demand for translation, and expanded our list of keywords to detect more translation-related user reviews,” Demid said. “[They] tested a wide range of natural language processing algorithms, including deep machine learning algorithms, to build a robust model that accurately predicts if a review praises, criticizes or requests the game’s translation.”

In the end, the result of the delivered work exceeded Demid’s expectations:

One student trained a model that demonstrated 98% precision at detecting new language requests — an outstanding result that was way beyond our initial expectations,” he said. “Two students delivered compiled Python scripts with the relevant user interface that we can use off-the-shelf!

Students’ feedback

This experience was also very beneficial for the students, who had their first opportunity to prove themselves in a real-life assignment.

“I was keen to be part of any real-world project and had been asking my community manager for such opportunities,” said TripleTen student Xia Cui. “So when I found out about the Allcorrect project, I signed up immediately. For me, any project would’ve been a good project to work on".

I learned so much from it. Not just about new packages and methods of analysis and reporting but also workflow management, how to clarify task requirements, collaborate with others, and more. Those are the experiences I can put on my resume, and that's awesome too.

With TripleTen externships, you will get hands-on experience at real tech companies that will become a solid basis for your portfolio and help you stand out during the job application process. To apply for one, enroll in one of the bootcamps and reach out to your community manager once you’ve passed Sprint 10 and feel confident enough to test yourself in action.