Live streaming recommender system is specifically designed to recommend real-time live streaming of interest to users. Due to the dynamic changes of live content, improving the timeliness of the live streaming recommender system is a critical problem. Intuitively, the timeliness of the data determines the upper bound of the timeliness that models can learn. However, none of the previous works addresses the timeliness problem of the live streaming recommender system from the perspective of data stream design. Employing the conventional fixed window data stream paradigm introduces a trade-off dilemma between labeling accuracy and timeliness. In this paper, we propose a new data stream design paradigm, dubbed Sliver, that addresses the timeliness and accuracy problem of labels by reducing the window size and implementing a sliding window correspondingly. Meanwhile, we propose a time-sensitive re-reco strategy reducing the latency between request and impression to improve the timeliness of the recommendation service and features by periodically requesting the recommendation service. To demonstrate the effectiveness of our approach, we conduct offline experiments on a multi-task live streaming dataset with labeling timestamps collected from the Kuaishou live streaming platform. Experimental results demonstrate that Sliver outperforms two fixed-window data streams with varying window sizes across all targets in four typical multi-task recommendation models. Furthermore, we deployed Sliver on the Kuaishou live streaming platform. Results of the online A/B test show a significant improvement in click-through rate (CTR), and new follow number (NFN), further validating the effectiveness of Sliver.

Adobe Fonts has a rich library of over 20,000 unique fonts that Adobe users utilize for creating graphics, posters, composites etc. Due to the nature of the large library, knowing what font to select can be a daunting task that requires a lot of experience. For most users in Adobe products, especially casual users of Adobe Express, this often means choosing the default font instead of utilizing the rich and diverse fonts available. In this work, we create an intentdriven system to provide contextual font recommendations to users to aid in their creative journey. Our system takes in multilingual text input and recommends suitable fonts based on the user's intent. Based on user entitlements, the mix of free and paid fonts is adjusted. The feature is currently used by millions of Adobe Express (shown in Figure 1) users with a CTR of >25%.

In this work we introduce an incremental learning framework for Click-Through-Rate (CTR) prediction and demonstrate its effectiveness for Taboola's massive-scale recommendation service. Our approach enables rapid capture of emerging trends through warm-starting from previously deployed models and fine tuning on "fresh" data only. Past knowledge is maintained via a teacher-student paradigm, where the teacher acts as a distillation technique, mitigating the catastrophic forgetting phenomenon. Our incremental learning framework enables significantly faster training and deployment cycles (x12 speedup). We demonstrate a consistent Revenue Per Mille (RPM) lift over multiple traffic segments and a significant CTR increase on newly introduced items.

Automated recommendations can nowadays be found on many e-commerce platforms, and such recommendations can create substantial value for consumers and providers. Often, however, not all recommendable items have the same profit margin, and providers might thus be tempted to promote items that maximize their profit. In the short run, consumers might accept non-optimal recommendations, but they may lose their trust in the long run. Ultimately, this leads to the problem of designing balanced recommendation strategies, which consider both consumer and provider value and lead to sustained business success. This work proposes a simulation framework based on agent-based modeling designed to help providers explore longitudinal dynamics of different recommendation strategies. In our model, consumer agents receive recommendations from providers, and the perceived quality of the recommendations influences the consumers' trust over time. We design several recommendation strategies which either give more weight on provider profit or on consumer utility. Our simulations show that a hybrid strategy that puts more weight on consumer utility but without ignoring profitability considerations leads to the highest cumulative profit in the long run. This hybrid strategy results in a profit increase of about 20 % compared to pure consumer or profit oriented strategies. We also find that social media can reinforce the observed phenomena. In case when consumers heavily rely on social media, the cumulative profit of the best strategy further increases. To ensure reproducibility and foster future research, we publicly share our flexible simulation framework.

Try here the demonstration tool for automatically classifying goods with their commercial descriptions and experience how AI could assist core Customs operations. As the awareness among Customs agencies about the importance and the interest in its application grows, the BACUDA expert team with the support of CCF-Korea continues to deliver state of the art methods and training material to meet the demands of Members. Complementing the development of the neural network model to support the classification of goods in Harmonized System, an online advanced Data Analytics course including a practical module on the HS recommendation algorithm was published on CLiKC!, the WCO e-learning platform. The BACUDA team of experts collaborated on the development of an AI model to recommend HS codes, which aims to support commodity classification for Customs officials by using historical data to predict HS codes upon the entry of the commercial descriptions of goods. An accompanying tool provides a demonstration on the functions which the model offers.

As it has in many other Industries, artificial intelligence has become a core component of e-commerce. Going forward, it's believed that AI and natural language understanding technologies in particular will find many new applications in the industry. E-commerce is a fast-growing industry. In 2017, global retail e-commerce sales reached US$2.3 trillion, and e-retail revenue is expected to grow to US$4.88 trillion in 2021. The world's continuous digitalization and informationization gives e-commerce a lot of room for development.

A couple of months ago I left Pivotal to join idealo.de Besides the usual tasks like building out the data science team, setting up the infrastructure and many more administrative stuff, I had to define the ML powered product roadmap. And associated with this was also the definition of a Minimum Viable Product (MVP) for machine learning products. The question I often face though, here at idealo and actually also back at my time at Pivotal, is what actually a good MVP means? In this article, I will shed some lights on the different dimensions of a good MVP for machine learning products drawing in the experiences that I've gained so far.

Machine learning is a collection of computer methodologies or algorithms that predict outcomes based on collating information from previous choices. These learning systems are adaptive, constantly evolving from new examples, and capable of determining internal parameters to recognize the nature of new data. ML acquires knowledge through the analysis of previous behaviors and/or experimental data, e.g. a learning dataset. Smart technology and AI applications and programs collect a vast amount of data which can then be analyzed to predict outcomes. Information obtained using machine learning methods are by far the most dependable way to predict results and construct reliable models, particularly if some data is unknown or unobtainable.

Many web search engines use the concept of a'spider' - automated software which goes out onto the web and trawls through the contents of each server it encounters, indexing documents as it finds them. This approach results in the kinds of databases maintained by services such as Alta Vista and Excite - huge indexes to a vast chunk of what's currently available on the web. However, the problems which users can face when using such databases are beginning to be well documented. A recent JISC-funded investigation [1] into the use of web search engines indicates that users can typically encounter a number of difficulties. These include the issue of finding information relevant to their needs, and the problem of information overload - when far too much information is returned from a search.