What are some ways we could use AI to improve our product?
- Better search. NLP can tag posts and questions with 'concepts'. Our search could be improved by matching concepts as well as keywords.
- Related discussions. There are a few algorithms to do this but maybe AI would do a better job. I imagine it would work in the same way as search.
- Summarize a discussion thread. I think this kind of thing might still be experimental.
- Auto-tagging. Maybe useful when you have a really big knowledge base.
Comments
-
- Better spam detection (may even help with ability to identify human spam)
- Suggested Rank criteria based on learned behaviours
- Suggested content before a discussion is created (reduce dupes)
- Recommendations about content
Here is how Salesforce uses their AI in Community (or how they see it being used/are marketing it)
https://www.salesforce.com/blog/2016/09/salesforce-einstein-smarter-productive-connections.html6 -
Adding meta data to uploaded content. That would allow you to do more advanced searches. An example would be searching for images of "people", without having explicit "people" tags on the uploaded images.
2 -
@slafleche said:
Adding meta data to uploaded content. That would allow you to do more advanced searches. An example would be searching for images of "people", without having explicit "people" tags on the uploaded images.I saw this Ted talk this weekend about this open source program that identifies things in pictures or video (in real time).
https://github.com/pjreddie/darknet
https://pjreddie.com/darknet/yolo/The software is general purpose to identify things, so out of the box, it doesn't know every type of object. However, you can train it to identify anything. For example, if we had a client with a car forum, they could upload images of different models to teach the AI to recognize those cars.
It knows how to identify people, but if we wanted to auto tag faces with usernames, we'd need samples of each user's face to do that.
I haven't actually tried it out, but the demo was pretty impressive.
0 -
Emotional analysis of posts. We could analyze text, emoji and images together to try and figure out if the discussion is mostly positive or negative. This could then be used as a tool for moderators or just feed it into the forum's analytics.
0 -
Acer has expressed very serious interest in Machine Translation tech. They feel that it's both inevitable and necessary for MT technology to breakdown the barriers of language and truly facilitated universal conversations for global customer communities.
0 -
I saw a talk at ConFoo about using AI for triaging support tickets and and answering some questions easy automatically. Then the user is asked if the answer they got fixes their issue. If not, then they get to talk to a human.
This could be a feature with a the knowledge base for answering FAQs.
0 -
http://www.businessinsider.com/sc/ai-is-improving-customer-service-experiences-heres-how-2018-3
"AVA's original catalog of knowledge came from chat logs, use cases, and forum posts. But as of today, AVA has analyzed more than 14 million sentences for keywords, entities, phrases, clusters, and other telling speech and language patterns."
0 -
I don't like approaching features from the side of "what can we do with this shiny tech", and I get super skeptical when folks / companies say they "have an interest" in a technology instead of a solution to a problem. When we talk about something like recommended discussions, that is a solved problem, we just don't have the bandwidth to implement it (even in the limited scope of the known solution). Adding AI is like throwing MongoDB into your tech stack to see what happens. Maybe it'll be 20% better performance but maybe it'll also mire you in a whole new category of bugs and problems you never knew existed and there's not sufficient docs to work thru.
When we previously tried using sentiment analytics our customers yawned and said it wasn't adding value. And that was before the vendor went bankrupt and left us in a lurch.
Google is heavy into this industry because their entire business model is basically having computers be as savvy as humans as a glorified complicated way to get you to click their marketing links. Our business model is to connect humans to other humans in a structured way. I'm dubious that having a computer middleman inhale & puke back out "meaning" is a good strategy to achieve that, and I'm super dubious the technology is at a point that it's useful even if we decide it meets our goals to move that way. Siri can't figure out how to set two alarms in the same spoken command (try "set alarms for 3pm and 4pm") and that bit of tech is produced by the highest-valued company on the planet.
0 -
@Lvez said:
http://www.businessinsider.com/sc/ai-is-improving-customer-service-experiences-heres-how-2018-3"AVA's original catalog of knowledge came from chat logs, use cases, and forum posts. But as of today, AVA has analyzed more than 14 million sentences for keywords, entities, phrases, clusters, and other telling speech and language patterns."
When I read something like that, I wonder if it's part of our role as a platform to provide that sort of processing, or if the companies are going to regard that as part of their "secret sauce" and something they need to do in-house to cross-reference against all support touch points. My take is that we should focus on being the best interface for both humans AND their backend services to connect to, to allow them to do that processing however they want. We may get to a point where we have a "NLP Processor" integration for a particular solution. In-housing the business logic to deal with particular customers' industries' support needs seems like a whole other extremely ambitious business plan.
0 -
I would focus where community managers are stuck doing grunt work that can be automated. Also, find a way to intelligently bring people back to the community when engagement shows signs of waning.
@Adrian said:
- Better spam detection (may even help with ability to identify human spam)
- Suggested Rank criteria based on learned behaviours
- Suggested content before a discussion is created (reduce dupes)
- Recommendations about content
Here is how Salesforce uses their AI in Community (or how they see it being used/are marketing it)
https://www.salesforce.com/blog/2016/09/salesforce-einstein-smarter-productive-connections.html1 -
I absolutely thinking watching for signs of fading engagement is something we should be investing in. I dunno if it requires AI to deduce that, but tracking onboarding progress and engagement level is something I've been advocating for since the beginning.
5 -
YouTube, the Great Radicalizer (New York Times)
It seems as if you are never “hard core” enough for YouTube’s recommendation algorithm. It promotes, recommends and disseminates videos in a manner that appears to constantly up the stakes. Given its billion or so users, YouTube may be one of the most powerful radicalizing instruments of the 21st century.
I think this is the insidious AI threat of our time, not some kind of SkyNet overlord. Building a system that blindly recommends things to optimize 1 metric is just stupidly dangerous.
0
