ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ •...
Transcript of ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ •...
![Page 1: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/1.jpg)
Content-‐Based Recommender Systems (CBR)
Vani Koduru Shankara Murthy
![Page 2: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/2.jpg)
u Introduction u Definition of CBR u Architecture u Advantages and Disadvantages u Future work
Agenda
![Page 3: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/3.jpg)
• Web is huge and growing day by day • Complex and time consuming to find an item of interest • Helps in making the decisions about the item of interest based
on preferences given to an item by the user • Most often used in e-‐commerce, movie, music and app
recommendations • Example-‐ web browsers, based on previous browsing history and
bookmarks • Three main filtering techniques adopted are • Content based Systems • Collaborative Systems • Hybrid Systems
What is a recommender system?
![Page 4: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/4.jpg)
![Page 5: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/5.jpg)
• Matches the attributes of the user profile against the attributes of a content object.
• Items are represented as set of structured data/features Movie; actor ,actress, director, year of release, rating, etc…
• Keyword-‐based Retrieval Models (VSM) • Parsing, stop words removal and stemming • tf-‐idf weighting , contents are represented as TDM • Basic string matching technique
• Semantic Analysis To address the problem of polysemy and Synonymy Interpret natural language documents
• Building the user profile • Monitoring and analyzing the user activity • Browsing history serves as the training data for machine learning
algorithms • Probabilistic Models
• CBR uses Cosine Similarity measure to find contents that matches user profile
What is CBR?
![Page 6: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/6.jpg)
High-‐level architecture
Source: State of the Art and Trends by Pasquale Lops, Marco de Gemmis, Giovanni Semeraro.
![Page 7: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/7.jpg)
Content Analyzer • Extracts the features (keywords, n-‐grams) from the source • Conversion from unstructured to structured item (item vector) • Data stored in the repository Represented Items Profile Learner • To build user profile • Updates the profile using the data in Feedback repository Filtering Component • Matching the user profile with the actual item to be recommended • Uses different Strategies • Cosine similarity
Architecture (contd…)
![Page 8: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/8.jpg)
Advantages * User Independence: Recommends only the items that interest
the user * Transparency: Recommendation is based on the item features,
explicitly list the contents features * New Item: Helps in recommending new items that are not yet
rated by other users
Disadvantages * Limited Content Analysis: They have the limit in the number and
type of the feature associated with an item * Overspecialization: Recommends those items that score high
with the user profile * Cold Start Problem: For a new user, systems don’t have
historical information to recommend items
Advantages and Disadvantages
![Page 9: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/9.jpg)
Challenges of feeding user with surprising and more reliable data Collaborative Filtering • Recommendations based on the ratings and preferences
given by other users • Cold Start Problem – requires initial data set
Hybrid Filtering • Combines the approaches of Content based and Collaborative
Filtering • Good Strategy • Amazon
Future Work
![Page 10: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/10.jpg)
References[1] http://facweb.cs.depaul.edu/mobasher/classes/ect584/Papers/ContentBasedRS.pdf
[2] Towards the next generation of recommender systems: A survey of state of arts and future extensions, Gadiminas, Advomavicius, member, IEEE, and Alexander Tuzhilin. Member, IEEE
[3] http://users.ics.forth.gr/~potamias/mlnia/paper_6.pdf���
[4] http://recommender-‐systems.org/content-‐based-‐filtering/ Cold start problem [5] http://ceur-‐ws.org/Vol-‐770/paper9.pdf
![Page 11: ContentBased+Recommender+ Systems+(CBR)€¦ · • Web+is+huge+and+growing+day+by+day+ • Complex+and+time+consuming+to+find+an+item+of+interest+ • Helps+in+making+the+decisions+about+the+item+of+interest+based+](https://reader033.fdocuments.us/reader033/viewer/2022060320/5f0d1b5f7e708231d438b64e/html5/thumbnails/11.jpg)
Thank you !