Voice Chat Discussion

First Riot Post
Comment below rating threshold, click here to show it.

Aeolian Melodies

Senior Member

06-27-2014

Quote:
Originally Posted by Lyte View Post
Like Amazon, Google, Facebook and other tech companies, we've developed in-house predictive models that can analyze chat logs in real-time and assess the level of certain sentiment. For example, the model can detect whether a chat log was generally positive communication, or generally negative communication, while also being able to specifically look at the frequency of certain words that we've labeled or curated as racist or homophobic (like f*g and f*ggot).
inb4 tons of people get wrongfully banned for joking around with friends in voice chat using the F and N words >.>


Comment below rating threshold, click here to show it.

Devi Freak Mint

This user has referred a friend to League of Legends, click for more information

Senior Member

06-27-2014

Quote:
Originally Posted by haaaaaaaalp View Post
Where did you get the data of all 5 strangers being on voice chat? Also, have you done a comparison of 5 players being on voice chat vs <5 players being on voice chat? The current system leads to most games having teams <5 players being on voice chat. It would be better for the game overall if voice chat was given to everyone so that teams all have 5 people on voice chat as opposed to 2-3 on voice chat.
How do you force 5 complete strangers to use voice chat?

Are we going to add owning a microphone and a headset to the minimum requirements?
Do we have to police people to make sure that they aren't blasting music over voice chat and game sounds?

As soon as one person opts out of voice communications, its usefulness drops dramatically
, you'll still have to use ping and text based comms.


Comment below rating threshold, click here to show it.

thunderingice

Senior Member

06-27-2014

Quote:
Originally Posted by Devi Freak Mint View Post
How do you force 5 complete strangers to use voice chat?

Are we going to add owning a microphone and a headset to the minimum requirements?
Do we have to police people to make sure that they aren't blasting music over voice chat and game sounds?

As soon as one person opts out of voice communications, its usefulness drops dramatically
, you'll still have to use ping and text based comms.
basic coercion. the second you don't use it and everyone else is, you become the guy who is bringing down the team.


Comment below rating threshold, click here to show it.

Frozt

This user has referred a friend to League of Legends, click for more information

The Council

06-27-2014

I've played many different games with different voice chat systems. DotA 2, CS:GO, Xbox Live, Rust/DayZ, etc. My opinion has always been that voice chat has significantly improved my experience in those games. Even if I'm being trashtalked, at least I understand, to a certain degree, who is trashtalking me. Text is always worse than voice.


Comment below rating threshold, click here to show it.

Ziodyne

Senior Member

06-27-2014

Quote:
Originally Posted by Devi Freak Mint View Post
How do you force 5 complete strangers to use voice chat?

Are we going to add owning a microphone and a headset to the minimum requirements?
Do we have to police people to make sure that they aren't blasting music over voice chat and game sounds?

As soon as one person opts out of voice communications, its usefulness drops dramatically
, you'll still have to use ping and text based comms.
The usefulness will drop, yes. But the coordination that voice chat gives is amazing, even if only 2 or 3 people are using.


Comment below rating threshold, click here to show it.

haaaaaaaalp

This user has referred a friend to League of Legends, click for more information

Senior Member

06-27-2014

Quote:
Originally Posted by Devi Freak Mint View Post
How do you force 5 complete strangers to use voice chat?

Are we going to add owning a microphone and a headset to the minimum requirements?
Do we have to police people to make sure that they aren't blasting music over voice chat and game sounds?

As soon as one person opts out of voice communications, its usefulness drops dramatically
, you'll still have to use ping and text based comms.
I'm not the one who wants to force 5 strangers onto voice chat. Lyte is CLAIMING that he has data from 5 strangers who were on voice chat (which is weird since voice chat with strangers isn't legal but...). I was questioning where that data came from and what they showed.


Comment below rating threshold, click here to show it.

Ding an Sich

Senior Member

06-27-2014

Lyte a question: what about other games that have implemented this feature? Such as Dota 2? Have you ever been able to talk to them or look at their working population?

Because all this information is really cool but how does it compare/contrast to other games?


Comment below rating threshold, click here to show it.

Lyte

This user has referred a friend to League of Legends, click for more information

Lead Social Systems Designer

Follow RiotLyte on Twitter

06-27-2014
14 of 17 Riot Posts

Quote:
Originally Posted by Ding an Sich View Post
Lyte a question: what about other games that have implemented this feature? Such as Dota 2? Have you ever been able to talk to them or look at their working population?

Because all this information is really cool but how does it compare/contrast to other games?
It's an interesting point. A few other studios who are making MOBAs have also decided to not implement voice chat, and have done similar research.

The study I mentioned did their research on Halo, but I imagine many of these findings apply across games/studios.


Comment below rating threshold, click here to show it.

Clubbems

This user has referred a friend to League of Legends, click for more information

Senior Member

06-27-2014

Quote:
Originally Posted by Lyte View Post
Which aspect are you curious about?

For part of the analysis you are referring to, we use a sentiment analysis model to review chat logs from a random sample of non-voice chat using players to a random sample of voice-chat using players. Then we break down the analysis in terms of premade sizes, wins versus losses, so on and so forth.
How do you know they were using voice chat? What voice chat were they using, etc.


Also, I have to recommend the way smite deals with voicechat. Them super emotes are amazing.


Comment below rating threshold, click here to show it.

Windstorm

Member

06-27-2014

Quote:
Originally Posted by eXigenT View Post
While what you say is true, the reason most studies use random sampling is because it is cheaper and less cumbersome than polling every single person in the community. However, since this is a video game, you are the one few places that can actually send a poll out to every single person with little effort. Makes me wonder why you don't just poll everyone?

At the end of the day, the reason for implementing voice chat is to improve overall communication and teamwork, right? I can say with certainty that using Curse Voice or voice chat in other games has improved my game experience lending to better teamwork. Of course you might get a few more cases of toxicity, but are we able to live with this possibility in order to improve our overall gaming experience? I think the answer is an emphatic yes--especially given the ability to mute toxic players in voice chat.
I've actually dealt with this problem before, and the answer isn't as easy to explain as it should be. essentially, at a basic level, the use of models and random sampling helps safeguard against seeing patterns when there are none. it is only in the last decade where we have tools that are able to handle large quantities of data reliably, and because the world is an imperfect place, sometimes correlations exist between things even when they are entirely separate or caused by a third variable.

in some ways its the stuff we learn in statistics 101, but on a scale that boggles the mind. take the classic example used to teach that correlation doesn't imply causation: ice cream sales and shark attacks. both increase during July and near beaches, and taken in isolation with bad methods, its "easy" for a person in a vacuum to reach the erroneous conclusion that they are linked (possibly with one causing the other). in fact, taken in context, we understand that they are both caused by the presence of more people at beaches during the summer.

large data sets present a similar problem, the tools we currently have are somewhat crude, and will spot every possible pattern in a given data set if it is large enough, even if it turns out to be useless or meaningless. use of random sampling and models acts as a kind of filter to remove some of the noise from those data sets. but just like trying to see an interleaved 3d image with a filter, it only works if you have the right one, and that's where a couple years of learning proper experimental design come in, so you can learn how to build the right filter.