This paper proposes using social tagging information from a recommender system to improve recommendations. It introduces two new tag-based recommendation techniques: social content-based filtering (SCBF) and social collaborative filtering (SCF). SCBF compares a user's tag cloud to item tag clouds, while SCF builds a target user tag cloud and compares it to other users' clouds. The techniques help address issues with standard collaborative filtering like cold start and provide more coverage. An example shows how SCBF could recommend a TV series based on a user's interest in surgery tags. The paper concludes the techniques provide more semantic connections through folksonomies and coverage.
International Journal on Natural Language Computing (IJNLC)kevig
?
We derive a method to enhance the evaluation for a text-based Emotion Aware Recommender that we have developed. However, we did not implement a suitable way to assess the top-N recommendations subjectively. In this study, we introduce an emotion-aware Pseudo Association Method to interconnect disjointed users across different datasets so data files can be combined to form a more extensive data file. Users with the same user IDs found in separate data files in the same dataset are often the same users. However, users with the same user ID may not be the same user across different datasets. We advocate an emotion aware Pseudo Association Method to associate users across different datasets. The approach interconnects users with different user IDs across different datasets through the most similar users' emotion vectors (UVECs). We found the method improved the evaluation process of assessing the top-N recommendations objectively
AN AFFECTIVE AWARE PSEUDO ASSOCIATION METHOD TO CONNECT DISJOINT USERS ACROSS...ijnlc
?
This document proposes an Affective Aware Pseudo Association Method to connect disjoint users across multiple datasets for validating a text-based emotion aware recommender system. The method uses user emotion vectors to associate users with different IDs across datasets based on the similarity of their emotion profiles. This allows combining data from different datasets to improve the evaluation process for assessing top recommendation lists. The document provides background on related work with emotion aware recommender systems and describes the methodology, which uses item and user emotion embeddings derived from sentiment analysis of text to build a multi-channel emotion aware recommender system.
l-Injection: Toward Effective CollaborativeFiltering Using Uninteresting ItemsJAYAPRAKASH JPINFOTECH
?
l-Injection: Toward Effective Collaborative Filtering Using Uninteresting Items
To buy this project in ONLINE, Contact:
Email: jpinfotechprojects@gmail.com,
Website: https://www.jpinfotech.org
The document discusses using spatial semantics and association rules to retrieve ocean images based on salinity and temperature variations over time. It proposes using concentric circles to annotate location information for objects in images and mining inter-transaction association rules to predict future trends. Experimental results demonstrate retrieving additional images based on salinity and temperature patterns to help users forecast variations.
This document summarizes a research paper that analyzes deep networks using kernel methods. It hypothesizes that (1) representations in higher layers of deep networks are simpler and more accurate, and (2) the network architecture controls how quickly representations are formed. The researchers used kernel principal component analysis to measure representation simplicity and accuracy at each layer of deep networks trained on MNIST and CIFAR. Their experiments found support for both hypotheses and that convolutional and pretrained networks form representations more systematically than standard multilayer perceptrons.
This document summarizes a study that used principal component analysis (PCA) and kernel principal component analysis (KPCA) to extract features from electrocardiogram (ECG) signals, which were then classified using a binary support vector machine (SVM) model. The study tested PCA, KPCA, and no feature extraction on ECG data from the MIT-BIH Arrhythmia Database to classify normal signals and three types of abnormalities. Results showed that combining SVM with KPCA feature extraction achieved the best classification performance compared to using SVM alone or with PCA. Automatic ECG classification is important for diagnosing cardiac irregularities.
Feature selection and classification in supporting report based self-manageme...es712
?
This document discusses using machine learning techniques to analyze self-reported data from people with chronic pain and identify their health status. It evaluates different feature selection methods and classification algorithms to determine an optimal approach for supporting self-management. The best performing method was found to be a multilayer perceptron classifier with high accuracy and area under the ROC curve, suggesting it could effectively classify health status levels from the self-reported data.
This document summarizes recent advances in collaborative filtering techniques for recommender systems. It describes how matrix factorization models have become popular for implementing collaborative filtering due to their accuracy. Neighborhood methods were also improved to be more accurate. The document outlines extensions that leverage temporal data and implicit feedback to further improve model accuracy. Key collaborative filtering approaches like matrix factorization, neighborhood methods, and techniques that combine their strengths are discussed.
An Experiment In Cross-Representation Mediation Of User ModelsDaniel Wachtel
?
This document presents the results of an experiment on cross-representation mediation of user models for movie recommendations. The experiment analyzed initializing user models for a content-based recommender using movie ratings from other sources. It tested inferring user preferences from ratings and determining default preferences from community ratings. The results showed this approach is feasible, with prediction errors decreasing for users with more ratings and stabilizing around 0.15. Community ratings best initialized an "average user model" to fill missing individual user values.
L injection toward effective collaborative filtering using uninteresting itemsKumar Dlk
?
We develop a novel framework, named as l-injection, to address the sparsity problem of recommender systems. By carefully injecting low values to a selected set of unrated user-item pairs in a user-item matrix secure computing in chennai
Content-based filtering with applications on tv viewing dataElaine Ceclia Gatto
?
This document discusses content-based filtering for recommender systems applied to TV viewing data in Brazil. It provides background on recommender systems and describes content-based filtering and related techniques like Apriori and cosine similarity that analyze item content correlations. It also briefly outlines Brazil's digital TV implementation and standards. The document aims to observe how content-based filtering can be used for recommendations in digital TV by testing TV viewing data.
A Research Paper on BFO and PSO Based Movie Recommendation System | J4RV4I1016Journal For Research
?
The objective of this work is to assess the utility of personalized recommendation system (PRS) in the field of movie recommendation using a new model based on neural network classification and hybrid optimization algorithm. We have used advantages of both the evolutionary optimization algorithms which are Particle swarm optimization (PSO) and Bacteria foraging optimization (BFO). In its implementation a NN classification model is used to obtain a movie recommendation which predict ratings of movie. Parameters or attributes on which movie ratings are dependent are supplied by user's demographic details and movie content information. The efficiency and accuracy of proposed method is verified by multiple experiments based on the Movie Lens benchmark dataset. Hybrid optimization algorithm selects best attributes from total supplied attributes of recommendation system and gives more accurate rating with less time taken. In present scenario movie database is becoming larger so we need an optimized recommendation system for better performance in terms of time and accuracy.
Content - Based Recommendations Enhanced with Collaborative InformationAlessandro Liparoti
?
This document discusses content-based recommender systems and describes a content-based collaborative (CBC) hybrid approach. CBC builds a content-based model enhanced with collaborative data by using a content-based similarity function that incorporates user-feature weights and feature importance weights. The CBC method is evaluated on movie recommendation datasets and is shown to outperform collaborative and content-based baselines, especially in new-item recommendation scenarios where no ratings exist.
Movie recommendation system using collaborative filtering system Mauryasuraj98
?
The document describes a mini project on building a movie recommendation system. It includes an abstract that discusses different recommendation approaches like demographic, content-based, and collaborative filtering. It also outlines the problem statement, proposed solution, workflow, dataset description, algorithm details, GUI design, result analysis, and applications. The system uses a user-based collaborative filtering model to recommend movies to users based on their preferences and ratings of similar users. Evaluation shows it has good prediction performance.
This document describes a movie recommendation engine that uses a hybrid approach combining content-based filtering and collaborative filtering. It first introduces recommendation systems and the different types, including content-based and collaborative filtering. It then outlines the steps to program the engine, including importing data, preprocessing it, fitting a KNN model, and displaying recommendations. The engine calculates similarity between movies to provide personalized recommendations to users based on their preferences.
Comparison of Collaborative Filtering Algorithms with Various Similarity Meas...IJCSEA Journal
?
Collaborative Filtering is generally used as a recommender system. There is enormous growth in the amount of data in web. These recommender systems help users to select products on the web, which is the most suitable for them. Collaborative filtering-systems collect users previous information about an item such as movies, music, ideas, and so on. For recommending the best item, there are many algorithms, which are based on different approaches. The most known algorithms are User-based and Item-based algorithms. Experiments show that Item-based algorithms give better results than User-based algorithms. The aim of this paper isto compare User-based and Item-based Collaborative Filtering Algorithms with many different similarity indexes with their accuracy and performance. We provide an approach to determine the best algorithm, which give the most accurate recommendation by using statistical accuracy metrics. The results are compared the User-based and Item-based algorithms with movie recommendation data set
Movie recommendation Engine using Artificial IntelligenceHarivamshi D
?
My Academic Major Project Movie Recommendation using Artificial Intelligence. We also developed a website named movie engine for the recommendation of movies.
COMPARISON OF COLLABORATIVE FILTERING ALGORITHMS WITH VARIOUS SIMILARITY MEAS...IJCSEA Journal
?
This document compares collaborative filtering algorithms with various similarity measures for movie recommendations. It summarizes User-based and Item-based collaborative filtering algorithms implemented in the Apache Mahout framework. Various similarity measures used in collaborative filtering are discussed, including Euclidean distance, Log Likelihood Ratio, Pearson correlation, Tanimoto coefficient, Uncentered Cosine, and Spearman correlation. The document concludes that Item-based algorithms typically provide better results than User-based algorithms for movie recommendations.
COMPARISON OF COLLABORATIVE FILTERING ALGORITHMS WITH VARIOUS SIMILARITY MEAS...IJCSEA Journal
?
Collaborative Filtering is generally used as a recommender system. There is enormous growth in the amount of data in web. These recommender systems help users to select products on the web, which is the most suitable for them. Collaborative filtering-systems collect users previous information about an item such as movies, music, ideas, and so on. For recommending the best item, there are many algorithms, which are based on different approaches. The most known algorithms are User-based and Item-based algorithms. Experiments show that Item-based algorithms give better results than User-based algorithms. The aim of this paper isto compare User-based and Item-based Collaborative Filtering Algorithms with many different similarity indexes with their accuracy and performance. We provide an approach to determine the best algorithm, which give the most accurate recommendation by using statistical accuracy metrics. The results are compared the User-based and Item-based algorithms with movie recommendation data set.
This document discusses analyzing a movie recommendation system using the MovieLens dataset. It compares user-based and item-based collaborative filtering approaches. For user-based filtering, it calculates user similarity using cosine similarity and predicts ratings. For item-based filtering, it also uses cosine similarity to find similar items and predicts ratings. It evaluates the performance of both approaches using root mean square deviation and finds that item-based collaborative filtering has lower error compared to user-based filtering.
This document provides an overview of recommender systems and different recommendation approaches, including content-based filtering, collaborative filtering using k-nearest neighbors, association rules, and matrix factorization. Collaborative filtering is described as the most widely used approach in practice and involves predicting a user's preferences based on the preferences of similar users. Matrix factorization techniques like singular value decomposition have gained popularity for modeling large, sparse user-item matrices in collaborative filtering.
Collaborative filtering is a technique used in recommender systems to predict a user's preferences based on other similar users' preferences. It involves collecting ratings data from users, calculating similarities between users or items, and making recommendations. Common approaches include user-user collaborative filtering, item-item collaborative filtering, and probabilistic matrix factorization. Recommender systems are evaluated both offline using metrics like MAE and RMSE, and through online user testing.
This document presents nearest bi-clusters collaborative filtering (NBCF), which improves upon traditional collaborative filtering approaches. NBCF uses biclustering to group users and items simultaneously, addressing the duality between them. It introduces a new similarity measure to achieve partial matches between users' preferences. The algorithm first performs biclustering on the training data. It then calculates similarity between a test user and biclusters to find the k-nearest biclusters. Finally, it generates recommendations by weighting items based on bicluster size and similarity. An example demonstrates how NBCF provides more accurate recommendations than one-sided approaches.
The document describes a multi-agent TV recommender system that uses three sources of user information - implicit viewing history, explicit preferences, and feedback on shows - to generate personalized program recommendations. It encapsulates this information into adaptive agents that collaborate to recommend shows. The system was tested on real users and found that the combination of implicit and explicit agents performed best.
Cervical cancer classification using gabor filters 1026es712
?
This document proposes using Gabor filters and K-means clustering to classify cervical biopsy images as normal, CIN1, CIN2, CIN3 or malignant. Images are preprocessed using Gabor filters to extract texture features, then segmented and classified using K-means clustering based on ratios of normal and abnormal cells. Evaluation shows this approach achieved sensitivities between 82-89% and specificity of 85% for cervical cancer grading.
A framework for emotion mining from text in online social networks(final)es712
?
This document proposes a framework for characterizing emotional interactions in social networks to distinguish friends from acquaintances. It collects posts and comments from social networks, develops lexicons to analyze informal language, generates features to assess text subjectivity, trains a model to classify text subjectivity, and uses this to train an SVM model that predicts relationships with 87% accuracy.
More Related Content
Similar to Exploiting social tagging in a web 2.0 recommender system(lab) (20)
This document summarizes recent advances in collaborative filtering techniques for recommender systems. It describes how matrix factorization models have become popular for implementing collaborative filtering due to their accuracy. Neighborhood methods were also improved to be more accurate. The document outlines extensions that leverage temporal data and implicit feedback to further improve model accuracy. Key collaborative filtering approaches like matrix factorization, neighborhood methods, and techniques that combine their strengths are discussed.
An Experiment In Cross-Representation Mediation Of User ModelsDaniel Wachtel
?
This document presents the results of an experiment on cross-representation mediation of user models for movie recommendations. The experiment analyzed initializing user models for a content-based recommender using movie ratings from other sources. It tested inferring user preferences from ratings and determining default preferences from community ratings. The results showed this approach is feasible, with prediction errors decreasing for users with more ratings and stabilizing around 0.15. Community ratings best initialized an "average user model" to fill missing individual user values.
L injection toward effective collaborative filtering using uninteresting itemsKumar Dlk
?
We develop a novel framework, named as l-injection, to address the sparsity problem of recommender systems. By carefully injecting low values to a selected set of unrated user-item pairs in a user-item matrix secure computing in chennai
Content-based filtering with applications on tv viewing dataElaine Ceclia Gatto
?
This document discusses content-based filtering for recommender systems applied to TV viewing data in Brazil. It provides background on recommender systems and describes content-based filtering and related techniques like Apriori and cosine similarity that analyze item content correlations. It also briefly outlines Brazil's digital TV implementation and standards. The document aims to observe how content-based filtering can be used for recommendations in digital TV by testing TV viewing data.
A Research Paper on BFO and PSO Based Movie Recommendation System | J4RV4I1016Journal For Research
?
The objective of this work is to assess the utility of personalized recommendation system (PRS) in the field of movie recommendation using a new model based on neural network classification and hybrid optimization algorithm. We have used advantages of both the evolutionary optimization algorithms which are Particle swarm optimization (PSO) and Bacteria foraging optimization (BFO). In its implementation a NN classification model is used to obtain a movie recommendation which predict ratings of movie. Parameters or attributes on which movie ratings are dependent are supplied by user's demographic details and movie content information. The efficiency and accuracy of proposed method is verified by multiple experiments based on the Movie Lens benchmark dataset. Hybrid optimization algorithm selects best attributes from total supplied attributes of recommendation system and gives more accurate rating with less time taken. In present scenario movie database is becoming larger so we need an optimized recommendation system for better performance in terms of time and accuracy.
Content - Based Recommendations Enhanced with Collaborative InformationAlessandro Liparoti
?
This document discusses content-based recommender systems and describes a content-based collaborative (CBC) hybrid approach. CBC builds a content-based model enhanced with collaborative data by using a content-based similarity function that incorporates user-feature weights and feature importance weights. The CBC method is evaluated on movie recommendation datasets and is shown to outperform collaborative and content-based baselines, especially in new-item recommendation scenarios where no ratings exist.
Movie recommendation system using collaborative filtering system Mauryasuraj98
?
The document describes a mini project on building a movie recommendation system. It includes an abstract that discusses different recommendation approaches like demographic, content-based, and collaborative filtering. It also outlines the problem statement, proposed solution, workflow, dataset description, algorithm details, GUI design, result analysis, and applications. The system uses a user-based collaborative filtering model to recommend movies to users based on their preferences and ratings of similar users. Evaluation shows it has good prediction performance.
This document describes a movie recommendation engine that uses a hybrid approach combining content-based filtering and collaborative filtering. It first introduces recommendation systems and the different types, including content-based and collaborative filtering. It then outlines the steps to program the engine, including importing data, preprocessing it, fitting a KNN model, and displaying recommendations. The engine calculates similarity between movies to provide personalized recommendations to users based on their preferences.
Comparison of Collaborative Filtering Algorithms with Various Similarity Meas...IJCSEA Journal
?
Collaborative Filtering is generally used as a recommender system. There is enormous growth in the amount of data in web. These recommender systems help users to select products on the web, which is the most suitable for them. Collaborative filtering-systems collect users previous information about an item such as movies, music, ideas, and so on. For recommending the best item, there are many algorithms, which are based on different approaches. The most known algorithms are User-based and Item-based algorithms. Experiments show that Item-based algorithms give better results than User-based algorithms. The aim of this paper isto compare User-based and Item-based Collaborative Filtering Algorithms with many different similarity indexes with their accuracy and performance. We provide an approach to determine the best algorithm, which give the most accurate recommendation by using statistical accuracy metrics. The results are compared the User-based and Item-based algorithms with movie recommendation data set
Movie recommendation Engine using Artificial IntelligenceHarivamshi D
?
My Academic Major Project Movie Recommendation using Artificial Intelligence. We also developed a website named movie engine for the recommendation of movies.
COMPARISON OF COLLABORATIVE FILTERING ALGORITHMS WITH VARIOUS SIMILARITY MEAS...IJCSEA Journal
?
This document compares collaborative filtering algorithms with various similarity measures for movie recommendations. It summarizes User-based and Item-based collaborative filtering algorithms implemented in the Apache Mahout framework. Various similarity measures used in collaborative filtering are discussed, including Euclidean distance, Log Likelihood Ratio, Pearson correlation, Tanimoto coefficient, Uncentered Cosine, and Spearman correlation. The document concludes that Item-based algorithms typically provide better results than User-based algorithms for movie recommendations.
COMPARISON OF COLLABORATIVE FILTERING ALGORITHMS WITH VARIOUS SIMILARITY MEAS...IJCSEA Journal
?
Collaborative Filtering is generally used as a recommender system. There is enormous growth in the amount of data in web. These recommender systems help users to select products on the web, which is the most suitable for them. Collaborative filtering-systems collect users previous information about an item such as movies, music, ideas, and so on. For recommending the best item, there are many algorithms, which are based on different approaches. The most known algorithms are User-based and Item-based algorithms. Experiments show that Item-based algorithms give better results than User-based algorithms. The aim of this paper isto compare User-based and Item-based Collaborative Filtering Algorithms with many different similarity indexes with their accuracy and performance. We provide an approach to determine the best algorithm, which give the most accurate recommendation by using statistical accuracy metrics. The results are compared the User-based and Item-based algorithms with movie recommendation data set.
This document discusses analyzing a movie recommendation system using the MovieLens dataset. It compares user-based and item-based collaborative filtering approaches. For user-based filtering, it calculates user similarity using cosine similarity and predicts ratings. For item-based filtering, it also uses cosine similarity to find similar items and predicts ratings. It evaluates the performance of both approaches using root mean square deviation and finds that item-based collaborative filtering has lower error compared to user-based filtering.
This document provides an overview of recommender systems and different recommendation approaches, including content-based filtering, collaborative filtering using k-nearest neighbors, association rules, and matrix factorization. Collaborative filtering is described as the most widely used approach in practice and involves predicting a user's preferences based on the preferences of similar users. Matrix factorization techniques like singular value decomposition have gained popularity for modeling large, sparse user-item matrices in collaborative filtering.
Collaborative filtering is a technique used in recommender systems to predict a user's preferences based on other similar users' preferences. It involves collecting ratings data from users, calculating similarities between users or items, and making recommendations. Common approaches include user-user collaborative filtering, item-item collaborative filtering, and probabilistic matrix factorization. Recommender systems are evaluated both offline using metrics like MAE and RMSE, and through online user testing.
This document presents nearest bi-clusters collaborative filtering (NBCF), which improves upon traditional collaborative filtering approaches. NBCF uses biclustering to group users and items simultaneously, addressing the duality between them. It introduces a new similarity measure to achieve partial matches between users' preferences. The algorithm first performs biclustering on the training data. It then calculates similarity between a test user and biclusters to find the k-nearest biclusters. Finally, it generates recommendations by weighting items based on bicluster size and similarity. An example demonstrates how NBCF provides more accurate recommendations than one-sided approaches.
The document describes a multi-agent TV recommender system that uses three sources of user information - implicit viewing history, explicit preferences, and feedback on shows - to generate personalized program recommendations. It encapsulates this information into adaptive agents that collaborate to recommend shows. The system was tested on real users and found that the combination of implicit and explicit agents performed best.
Cervical cancer classification using gabor filters 1026es712
?
This document proposes using Gabor filters and K-means clustering to classify cervical biopsy images as normal, CIN1, CIN2, CIN3 or malignant. Images are preprocessed using Gabor filters to extract texture features, then segmented and classified using K-means clustering based on ratios of normal and abnormal cells. Evaluation shows this approach achieved sensitivities between 82-89% and specificity of 85% for cervical cancer grading.
A framework for emotion mining from text in online social networks(final)es712
?
This document proposes a framework for characterizing emotional interactions in social networks to distinguish friends from acquaintances. It collects posts and comments from social networks, develops lexicons to analyze informal language, generates features to assess text subjectivity, trains a model to classify text subjectivity, and uses this to train an SVM model that predicts relationships with 87% accuracy.
This document presents a method for classifying road environments using images from a vehicle-mounted camera. It extracts color and texture features from subregions of road images and classifies them using k-NN and artificial neural networks (ANN). For a four-class problem distinguishing off-road, urban, major road, and motorway classes, the accuracy is around 80%. For a two-class problem distinguishing off-road and on-road, the accuracy increases to around 90% using ANN classification. The method achieves a near real-time classification rate of 1Hz by classifying one video frame per second.
Classification of commercial and personal profiles on my spacees712
?
This document presents research on classifying profiles on MySpace as either commercial or personal. A decision tree classifier called J48 was developed that uses profile attributes such as gender, age, friends, and publishing relationships. The J48 classifier achieved an accuracy of 92.25% to 96.42% on test data. The classifier is then applied to a privacy-preserving system where personal profiles can publish anonymously through an "avatar" to avoid disclosing private information to commercial profiles.
Tennis video shot classification based on support vectores712
?
This document proposes a tennis video shot classification method based on support vector machines. It extracts edge distribution, optical flow, and shot classification features. Optical flow features include foreground tracked point ratio and mean motion vector length. These 7 features are input to a multi-class SVM classifier to categorize shots as long shot, audience shot, close shot, or close-up shot. An experiment applies this approach to tennis videos and achieves higher accuracy than methods based solely on color distribution.
Social media recommendation based on people and tags (final)es712
?
1) The document proposes methods to generate personalized recommendations in social media platforms based on people relationships and tags.
2) An evaluation of three recommendation approaches that utilize direct tags, indirect tags through related items, and incoming tags from other users found that a combination of direct tags and incoming tags most accurately represented a user's interests.
3) A user study tested five recommendation approaches and found that combining people relationships and tags into a user profile achieved the highest ratings for interesting recommendations and lowest for non-interesting items.
Social media recommendation based on people and tags (final)es712
?
Exploiting social tagging in a web 2.0 recommender system(lab)
1. This paper appears in:
Internet Computing, IEEE
Date of Publication: Nov.-Dec. 2010
Product Type: Journals & Magazines
Exploiting Social Tagging in a Web 2.0
Recommender System
Ana Beln Barragns-Martnez
Centro Universitario de la Defensa en la Escuela Naval Militar de Marn, Spain
Marta Rey-Lpez
Consellera de Educacin e O.U., Spain
Enrique Costa-Montenegro, Fernando A. Mikic-Fonte, Juan C. Burguillo, and
Ana Peleteiro
University of Vigo, Spain
Student: Chen-Ting Huang
Advisor: Yin-Fu Huang
2. Issues
To take advantage of Web 2.0 applications, the
authors propose using information obtained from
social tagging to improve recommendations.
The Web 2.0 TV program recommender queveo.tv
currently combines content-based and collaborative
filtering techniques.
This article presents a novel tag-based recommender
to enhance the recommending engine by improving
the coverage and diversity of the suggestions.
3. Motivations
The ratings are related to the tags rather than to the
items themselves, which makes them valuable even
when the items are no longer in the system.
The hybrid proposal works well because the algorithms
complement each other; CBF(content-based filtering)
and CF(collaborative filtering) recommends.
This lets us enrich our recommender system with two
new recommendation techniques, SCF and SCBF, both
based on social tagging information.
4. Social tagging and Folksonomy
we propose taking advantage of the systems social
tagging capabilities to enrich the quality of the
recommendations.
Social tagging also lets us create a folksonomy,
which
shows the relationships between the different tags.
Such an approach would improve the quality of
recommendations twofold.
5. Recommender Systems
The standard CF approach presents some well-known
problems:
? gray-sheep problem
? cold-start problem
? first-rater problem
The primary drawback to CBF systems is their
tendency to overspecialize item selection .
We adopt a hybrid approach(based on social tagging)
for the TV recommender domain.
7. Tag-Based Recommenders
queveo.tv lets users give items tags to describe
them.
Use these tags to build both User and Item tag
clouds.
The weight of the tags is proportional to the number
of times they have been assigned
In user tag clouds, a tags weight is also proportional
to the ratings the users gave the items.
7
8. User tag cloud
In user tag clouds, a tags weight is also proportional
to the ratings the users gave the items.
User clouds consist of tags users have never assigned.
User tag cloud
8
9. Item tag cloud
Item tag cloud includes the tags users have assigned
to
it.
Item tag clouds reflect the relationships between the
systems tags. We represent this structure, called a
folksonomy. Item tag cloud
9
10. Social content-based recommendations
The simplest way of recommending items to users is
by directly comparing their tag clouds.
We measure the number of coincident tags of both tag
clouds (direct relationship, R0)
?The weight of the tags is proportional to the number of times they
have been assigned
?In user tag clouds, a tags weight is also proportional to the ratings
the users gave the items.
12. Social collaborative recommendations
A new tag cloud for the items, called the target-user tag
cloud
The system compares it with the potential users tag
clouds to obtain their similarity
13. Illustrative Example - CBF
A new user Juan is entering queveo.tv. He selects the
categories Documentaries: General and Medical and
Sports: Basketball as his likes.
? The CBF algorithms output consists of those TV programs that match his
likes More Than a Game and NBA Action
? Recommends another documentary, The Operation: Surgery Live.
CBF recommendation
14. Illustrative Example - CF
The main goal of our item-based CF approach is to
precisely fill these empty values with predictions.
For a recommendation threshold of seven, the CF
algorithm recommended the documentary and the TV
series.
CF(item-based) recommendation
14
15. Illustrative Example - CF
More Than a Game was rated with the same pattern as
The Operation: Surgery Live that is, both Marta and
Fernando gave a similar rating to both.
15
16. Illustrative Example - CF
Enrique and Fernando gave similar ratings to NBA Action
and House
16
17. Illustrative Example - CF
Without the tag-based recommender, the final results
would be The Operation: Surgery Live (both CF and CBF
recommend it) and House.
CF recommendation
17
18. Illustrative Example - SCBF
We compose the provisional user tag cloud, which the
system uses until users have their own.
The system can now use the information from both Juans
tag cloud and the tag clouds from each program.
18
19. Illustrative Example - SCBF
Juan has used terms such as surgery and doctor to tag the
documentary The Operation: Surgery Live.
The SCBF algorithm finds new relevant content to
recommend: the TV series Nip/Tuck (focused on a plastic
surgery practice).
19
20. Illustrative Example - SCF
the SCF algorithm also recommends to Juan the movie
Apollo 13.
Because its target-user tag cloud contains tags that are
also in Juans tag cloud or related to them through the
folksonomy.
By SCF By SCBF
21. Conclusions and Future Work
Using tag-based recommendation techniques lets
queveo.tv gain more semantic interconnections
thanks to the use of folksonomies.
They also obtain greater coverage because additional
relevant items are now included among the
recommendations .
In the future, we will study the possibility of reducing
the weights of the tags as they get older .