Session 5 Flashcards
What is a recommender system?
show in practice when we open online shops and recommend us quite suitable products
what are benfits of a RS to customer?
get better suitable recommendations (reduce search costs), be able to discover new products
what are benefits of RS to the seller?
cross-selling opportunities to make more money, higher customer satisfaction
what is collaborative filtering?
filtering based on users behaviour and similar decisions made by other uses
–> we look for a group of similar users and check whether they bought this product, if yes we recommend it to the new users (“others who bought this book, also bought these”)
- item and user based filtering
what is content-based filtering?
filtering based on content
-> if we have a book with similar content we recommend it to people who bought the similar book
what are challenges with content based filtering?
- suggests the same content every time
- often the start, as you need a growd for collaborative filtering (item and user based)
what are challenges /problems to overall recommender systems?
- cold-start problem for new users and new items
- scalability: calculations become difficult, when business is growing, because of more data
- sparsity problem: sometimes difficult to find users with the same prachsing history
- exotic profiles: make it hard to make suitable recommendations (when more than 1 person uses account)
what is a solution to the cold-start problem?
use APIs ro conncet online shops withf Facebook, which has a lot of information and can be integrated into the recommendations (raises privacy concerns)
how does collaborative filtering: user-based nearest neighbour recommendation work?
if users had similar tastes in the past, they will also in future: user preferences remain stable -> generate ratings for everyon unseen/ purchased item
how does collaborative filtering: item-based nearest neighbour recommendation work?
we can also calculate similarieties between items, this is favoured in practie
-> based on ratings of users (If you rate “Toy Story” and “Finding Nemo” highly, the system might recommend “Monsters, Inc.” because its rating pattern is similar to “Finding Nemo” and “Toy Story” based on other users’ ratings)
how does content-based filtering work?
recommendation task consist of determining the items that match the user’s preferences best
does not reuire large user base or rating history, but knowlegdge of item characteristics
what is a chatbot?
a computerized service that enables easy conversation between humans and humanlike computerizted robots or image characters
what are the three main catrgoies of chatbot applications?
- enterprise chatbots
- virtual personal assistance (siri, Alexa, etc.)
- robo advisors
how does a chatbot work?
how cann DSS work as a judge advisor system?
a judge (us, decision-maker) decised about an afvisor (human, algorithm, human-algrothm team) giving advice (chatgpt, output, recommendation favrouring or discouraging particular options)
How does this judge advisor system look in real life (example and model)?
- Example doctor: trust à do we trust the doctor?; accuracy (is the advice helpful?); advice utilization ( do we take the recommended number of pills?), system satisfaction ( are we happy with the entire interaction with the system?); intention to continue/use (do we want to continue using X)
- Framework originates from human-to-human interaction but is now used for human-to-machine interaction
