Yelp collects and hosts reviews and ratings for all things local. Over years it has collected zillions of such records. Any one in the internet services industry understands that it is like a diamond mine. However, I see quite a few issues with their search. The site is relatively slow compared to many other sites. The local search relevancy is not as good as Google. Hence, it is often easier to find specific local places in yelp by first starting the search in Google. They recently made a few new features such as detection of interesting phrases etc. These NLP features are good but I believe they have only touched the surface of their potential.
Sunday, September 9, 2012
Saturday, September 8, 2012
Pandora's and Apple
Pandora's business model is based on advertisements that they throw when someone is listening to their radio service. They have both audio and video advertisements. The video ads are quite ineffective because the person listening to the music may not be watching the display of the device. The audio ads are most of the time quite obtrusive and it seems they dont have any concept of relevance while placing the audio ads. Hence, the technology and business of ad supported model is not without flaws. They also are soon going to be subjected to new competition from Apple. Apple already has about 400 Million or more iTune users. It is simpler for them to integrate the pandora like radio seamlessly with every IOS devices. Apple may prefer a subscription model over ad supported model. They also already has a social service called "ping". Overall, Apple no doubt has a better chance in getting most of the market for the collaborative radio service. Pandora should start thinking about 1. business model 2. technology that can support their business model. In my estimation they may get about 2 years time before the Apple gets into the market fully. Hence, Pandora still have time to innovate.
Friday, September 7, 2012
Image Quality in Online Shopping
It turns out that the perceptual quality of the product images is an important factor for improving customer engagement in e-commerce sites. I have recently published two papers with some of my current team members on studies of impact of image quality in online shopping ( WWW 2011, MDMKDD 2012). In these papers, we have shown how one can use an image quality measure to improve the NDCG of clicks in online shopping sites. This aspect was also studied by marketing researchers in past since it sounds like a no brainer that images are important for online shopping.
However, I believe it can actually be more important in future than it is today with the advent of new technologies in digital photography, availability of high quality camera with smart phones. The assessment of image quality is a hard problem. In one of the papers I wrote on this topic ( SPIE 2012), I have discussed the notion of image quality in the context of online shopping. However, a lot more work can be done in this area to use high level visual attributes in the assessment of image quality.
However, I believe it can actually be more important in future than it is today with the advent of new technologies in digital photography, availability of high quality camera with smart phones. The assessment of image quality is a hard problem. In one of the papers I wrote on this topic ( SPIE 2012), I have discussed the notion of image quality in the context of online shopping. However, a lot more work can be done in this area to use high level visual attributes in the assessment of image quality.
Thursday, September 6, 2012
Social shopping
The core idea of social shopping is that we often buy things based on our friends recommendations. The recommendation can be explicit or implicit. In an explicit recommendation, a friend proactively recommends a product to another friend. In an implicit recommendation, a friend broadcasts his or her recent purchases which may trigger a purchase interest to some of his/her friends. Booz & Co released a study estimating that the dollar value of goods sold through social media will reach $30Bil worldwide ($14Bil in the US) by 2015 up from $5Bil worldwide ($1Bil in the US) in 2011. Business Insider published a report estimating that the flash sales segment (e.g. Gilt, Fab, One Kings Lane, ideeli, Totsy, etc.) of social commerce will reach $6Bil by 2015 up from an estimated $1.5Bil in 2011. However, so far the notion of right way to implement a social shopping infrastructure has not moved far beyond posting and broadcasting one's purchases in Facebook.
Tuesday, June 19, 2012
Broad queries in Amazon's product search
In web search sometimes users search with a query which covers a broad set of concepts or categories. In the context of e-commerce, such queries may cover a broad range of products. Typically, a search engines best strategy to deal with such broad queries can be to provide a diverse set of results optimally covering these various concepts or categories. There are no dearth of academic studies on classification of queries. I have recently noticed that Amazon's product search is not giving very good results for quite a few popular broad queries.
Here are a few example:
Query: Mac
Query: Foundation
Here are a few example:
Query: Mac
Query: Foundation
Wednesday, August 31, 2011
On Machine Learning
The discipline of Machine Learning has inherited a major portion of it's theory from classical statistics. Today, the area is quite vast and it consists of many topics including classification, regression, clustering, reinforcement learning etc. We often see interview candidates with machine learning in resume. In academic circle, typically a researcher in machine learning is expected to publish in ICML, NIPS, COLT community.
Application research can be published in various different conferences such as KDD, ICDM, SDM, CVPR, ECCV, ICPR, SIGIR and even in WWW. However, the question arises that what knowledge is really required or valuable for software engineering positions. I feel that someone claiming some practical and academic knowledge in machine learning should be able to answer some of the questions such as follows:
1. What are sample complexity and model complexity?
2. What are the characteristics of feature selection algorithms?
3. How one handle sparsity in features?
4. What are various loss functions used for SVM, clustering, decision trees?
5. What are the desirable characteristics of loss functions and why?
6. What is regularization?
7. Why generalized linear models are called "generalized"?
8. What are exponential family distributions? What is their significance in machine learning?
9. Can one make a classifier using regression model?
10. How do one compare various models with same target and with different targets in regression?
11. Can one apply k-means algorithms for a data set with Zipf distribution? What modification is required?
12. What is Gibb's sampling?
13. What is the difference between HMM and CRF?
14. What is the complexity of inference in Graphical model?
The above questions are a few of the sample questions. I believe a good reading of selected chapters of Duda and Hurt, Bishop and Hastie's book (3 books total) can provide one a basic background in the subject. Boyd's convex optimization book and Feller's volume one in statistics can be used as references for knowledge in convex optimization and statistics.
Application research can be published in various different conferences such as KDD, ICDM, SDM, CVPR, ECCV, ICPR, SIGIR and even in WWW. However, the question arises that what knowledge is really required or valuable for software engineering positions. I feel that someone claiming some practical and academic knowledge in machine learning should be able to answer some of the questions such as follows:
1. What are sample complexity and model complexity?
2. What are the characteristics of feature selection algorithms?
3. How one handle sparsity in features?
4. What are various loss functions used for SVM, clustering, decision trees?
5. What are the desirable characteristics of loss functions and why?
6. What is regularization?
7. Why generalized linear models are called "generalized"?
8. What are exponential family distributions? What is their significance in machine learning?
9. Can one make a classifier using regression model?
10. How do one compare various models with same target and with different targets in regression?
11. Can one apply k-means algorithms for a data set with Zipf distribution? What modification is required?
12. What is Gibb's sampling?
13. What is the difference between HMM and CRF?
14. What is the complexity of inference in Graphical model?
The above questions are a few of the sample questions. I believe a good reading of selected chapters of Duda and Hurt, Bishop and Hastie's book (3 books total) can provide one a basic background in the subject. Boyd's convex optimization book and Feller's volume one in statistics can be used as references for knowledge in convex optimization and statistics.
Sunday, July 31, 2011
An analysis of MSFT's search war strategy
Comscore report published that Bing combined with Yahoo has now 41% search market share. This result is doubtful. However, it is possible for Bing to get increased search share from Facebook queries.
The MSFT's search strategy began with an ambition to make another search engine like Goog. I believe that Gates thought about using their PC domination, friend company network in gaining market from Goog exactly the way they killed Netscape. However, Goog rises into a giant business with increasing advertising dollar.
In phase one of the search war, MSFT had their own loyal engineers working in search. But, the leadership soon realized that search has more "algorithms" and "AI" and it is development of "distributed systems" and internet services which is certainly different that PC based software applications. In second phase, MSFT started following Goog's recruitment pattern hiring PhDs and scientists. They also replaced the position of head of search with someone who has more academic background in IR. However, the strategy of making a better Google search remained same. Goog at the same time made better email service, a better translation service and many other new services that MSFT leaders could not think ahead of time. In the third phase, MSFT leaders realizes that Goog has a platform and two sided effect has already made them very powerful and it is almost impossible to gain search market share without innovation. Traditionally innovation is weakest in MSFT's product development. They then hired veterans from falling Yahoo to run their search business and they started making "Bing" as a decision engine. This is a new phase for MSFT's search and also it is a new generation for MSFT's strategy. They still lack an out of the box original something that can make Bing attractive choice for search and unfortunately, this is hard unless MSFT makes it strategically important to make innovation in search and can spread across their org. An example of such innovation by MSFT is "kinnect" and "xbox live".
The MSFT's search strategy began with an ambition to make another search engine like Goog. I believe that Gates thought about using their PC domination, friend company network in gaining market from Goog exactly the way they killed Netscape. However, Goog rises into a giant business with increasing advertising dollar.
In phase one of the search war, MSFT had their own loyal engineers working in search. But, the leadership soon realized that search has more "algorithms" and "AI" and it is development of "distributed systems" and internet services which is certainly different that PC based software applications. In second phase, MSFT started following Goog's recruitment pattern hiring PhDs and scientists. They also replaced the position of head of search with someone who has more academic background in IR. However, the strategy of making a better Google search remained same. Goog at the same time made better email service, a better translation service and many other new services that MSFT leaders could not think ahead of time. In the third phase, MSFT leaders realizes that Goog has a platform and two sided effect has already made them very powerful and it is almost impossible to gain search market share without innovation. Traditionally innovation is weakest in MSFT's product development. They then hired veterans from falling Yahoo to run their search business and they started making "Bing" as a decision engine. This is a new phase for MSFT's search and also it is a new generation for MSFT's strategy. They still lack an out of the box original something that can make Bing attractive choice for search and unfortunately, this is hard unless MSFT makes it strategically important to make innovation in search and can spread across their org. An example of such innovation by MSFT is "kinnect" and "xbox live".
Subscribe to:
Posts (Atom)
