Современные подходы к проблемам в рекомендательных системах Александр Тужилин New York University, Сбербанк AI Lab
Outline of the Talk • Traditional matrix completion paradigm of recommender systems • Going beyond the traditional matrix completion paradigm • Example of a non-traditional approach – recommending products with the most valuable aspects based on customer reviews [KDD 2017] 2
Recommender Systems (RS) 3
The Traditional Paradigm of Recommender Systems … s 1 s 2 s N • Matrix R of known ratings c 1 ˆ – r ij : rating user c i assigns to item s j c M R R c 2 f ( R , X , Y ) • Matrix X of user attributes … – x ij : attribute x j of user c i • Matrix Y of item attributes … … x 1 x 2 x P s 1 s 2 s N – y ij : attribute y j of item s i c 1 c 1 ˆ c M X c 2 • A matrix completion problem : c 2 c M R … … estimate unknown ratings in 𝑆 Solutions [AT05]: … y Q y 1 y 2 • Collaborative Filtering (CF) s 1 • Content-Based s N Y s 2 • Hybrid … Example: Netflix Prize Competition 4
Netflix Competition Recommendation Matrix K-PAX Life of Memento Notorious Brian U1 4 3 2 4 U2 4 5 5 U3 2 2 4 U4 3 5 2 • The Users × Items Matrix of Ratings (w/ timestamps) • Key issue: accurate estimation of unknown ratings (RMSE) 5
Characteristics of the Traditional (Matrix Completion) Paradigm • Two-dimensional ( 2D ) paradigm: Users and Items – 3 matrices: R, X and Y • Utility of an item to a user revealed by a single rating – binary or multi-scaled • Recommendations of individual items provided to individual users – Many solution via estimation of unknown ratings 6
Going Beyond the Traditional Matrix Completion Paradigm Remove limiting assumptions of the traditional paradigm: • Go beyond the characteristics of (a) User × Item matrix, (b) single ratings, (c) recommending individual items to individual users [ JRTZ16] “Recommender Systems —Beyond Matrix Completion” New performance Traditional time 7
Going Beyond the Traditional Matrix Completion Paradigm • Context-Aware RSes (CARS) – Including spatiotemporal and mobile RSes • Multi-criteria ratings • Aggregate ratings and recommendations to groups • Flexible and constraint-based recommendations • Non-rating-based approaches (e.g., ranking-based) • New performance measures: novelty, serendipity, … • Social RSes • User interactions/feedback, e.g., “conversational” RSes • Trust and privacy • Manipulation-resistant RSes • Additional data sources, such as customer reviews 8
Recommending Products with Valuable Aspects Based on Customer Reviews Joint work with K. Bauman (Temple) and B. Liu (UIC) 9
Research Idea Idea: Recommending not only a product but also the most important (positive or negative) aspects that can enhance customer experience with the product. Examples: Positive: visit “ Aquagrill ” and order “FISH” there. Negative: visit “Cafe X” but do not order “DESSERT” there . Aspects come from customer reviews (e.g. Yelp). 10
Recommending Products and Aspects Try Goat Cheese Try Sweet Mango Sauce Try Lamb Curry or Chicken Masala Avoid Thai Tea 11
Importance of such Recommendations Why is it important? Novel approach to RSes that provides more tangible recommendations that enhance customer experience with the products. 12
Method of Recommending the Most Valuable Aspects Input: set of historical reviews with ratings. Output: product recommendations with the most valuable aspects enhancing customer experiences. 13
Method of Recommending the Most Valuable Aspects: An Overview 1. Extracting aspects from the reviews 2. Training Sentiment Utility Logistic Model (SULM) • aspect sentiments • overall satisfaction 3. Calculating aspect impact on rating 4. Recommending products and aspects 14
1. Extracting Aspects and Sentiments from the Reviews • Determine set A of aspects in an application • Aspects are characterized by a set of terms, e.g. MEAT - {meat, pork, bbq, lamb, ..} • For each review r determine set of aspects A r discussed in r and corresponding set of sentiments • Use Opinion Parser [Liu, 2010]. 15
1. Extracting Aspects and Sentiments from the Reviews Example: “( 1) Had lunch in Taqueria today. (2) Ordered the taco with rice and beans and it was great . (3) The service was quick. (4) The atmosphere was dark and soothing .” • FOOD – positive • SERVICE – positive • ATMOSPHERE – positive 16
2. Sentiment Utility Logistic Model (SULM) • Main purpose is to estimate the overall customer experience • Simultaneously fits ratings and sentiments • Identifies relative importance of the aspects in the potential customer experience 17
2. (a) Training SULM: Aspect Sentiments Expressed sentiment (OP output): - level of satisfaction Sentiment utility of user with aspect of item , latent variable. Use Matrix Factorization (Koren et al. 2009) to estimate: 18
2. (b) Training SULM: Explicit Sentiment vs. Actual Value of an Aspect 1 Logistic function: Estimation of explicit sentiment: 0,5 0 Maximize log-likelihood: -6 -4 -2 0 2 4 6 s so that estimated values of sentiments fit the real binary Estimate 𝜄 sentiments o k u i extracted by OP 19
2. (c) Training SULM: Overall Satisfaction Overall rating (binary): Overall level of satisfaction (latent): 20
2. (d) Training SULM: Explicit Rating vs. Actual Value of an Experience 1 Binary rating estimation: 0,5 0 Maximize log-likelihood: -6 -4 -2 0 2 4 6 21
2. Scheme of SULM Estimate parameters of SULM such that 1 1) 0,5 0 -6 -4 -2 0 2 4 6 2) 1 0,5 0 -6 -4 -2 0 2 4 6 22
2. Training SULM Simultaneously fits ratings and sentiments: Use Stochastic Gradient Descent to fit the model. 23
3. Aspect Impact on Rating For a new potential review compute the impact of each aspect on the predicted rating as the corresponding summand from the rating prediction part of the model 24
4. Recommending Products and Aspects • identify group of aspects over which the customer has control (e.g. Dish vs. Decor); same for the management • identify the most valuable aspects of the potential experience • recommend a product and its corresponding suggestions to experience (positive) or do not experience (negative) a particular aspect 25
4. Recommending Aspects to Managers Provide complimentary drink Recommend foot massage Do not talk too much during the procedure Manager of a Spa Salon 26
Experimental Settings Restaurants Hotels Beauty&Spa Initial 1,344,405 96,384 104,199 Filtered 602,112 5,669 5,065 Users 23,209 352 349 27
Examples of Aspects (Restaurants) Meat Fish Dessert Money Service Decor beef cod tiramisu price bartender design meat salmon cheesecake dollars waiter ceiling bbq catfish chocolate cost service décor ribs tuna dessert budget hostess lounge veal shark ice cream charge manager window pork fish macaroons check staff space 28
Numbers of Aspects Restaurant Hotel Beauty & Spa 65 42 45 Total number Customer can 49 14 17 control Management can 54 29 31 control 29
Performance Measures 1. Recommendations of Aspects : how much the average rating is changed for those customers who followed the recommendations of aspects 2. Rating prediction : AUC, Precision@Top3 3. Aspect ranking : Precision@Top3, Precision@Top5 30
Baselines • Recommending the most popular aspect of a product • Recommending highly rated aspect of a product • Hidden Factors as Topics ( HFT ) (McAuley and Leskovec 2013) - state-of- the-art rating prediction method based on customer reviews • Learning to Rank User Preferences Based on Phrase-Level Sentiment Analysis across Multiple Categories ( LRPPM ) (Chen et al. 2016) - the latest method for predicting the list of aspects appearing in the review 31
Recommendations of Aspects (Restaurants) 72.3% Followed Positive 65.1% Average 62.9% Not Followed Positive Conclusion: our recommendations help to get better customer experience as captured by the ratings 32
Recommendations of Aspects (Restaurants) Followed positive recommendations 74,0 % Average 72,0 % Most Popular 70,0 % Highly Rated 68,0 % SULM 66,0 % Conclusion: our 64,0 % recommendations help to get 62,0 % better customer experience vs. 60,0 % baselines. Customers Managers 33
Recommendations of Aspects (Restaurants) Did not follow positive recommendations 66,0 % Average Lowest sentiment 65,0 % SULM 64,0 % 63,0 % Conclusion: our recommendations help to avoid 62,0 % negative customer experience 61,0 % better than baselines Customers Managers 34
Recommendations of Aspects (Hotels) Followed positive recommendations 70,0 % Average Most Popular 65,0 % Highly Rated SULM 60,0 % Conclusion: our 55,0 % recommendations help to get better customer experience vs. 50,0 % baselines. Customers Managers 35
Recommendations of Aspects (Beauty & Spa) Followed positive recommendations 76,5 % Average 75,0 % Most Popular Highly Rated 73,5 % SULM 72,0 % Conclusion: our 70,5 % recommendations help to get 69,0 % better customer experience vs. 67,5 % baselines. Customers Managers 36
Recommend
More recommend
