New item or topic profiling and recommendation are useful yet challenging, especially in face of a “cold-start” situation with sparse user-item ratings for the new arrivals. In this paper, a method of acquiring review opinions of the “sentinel” users on the cold-start items is proposed to elicit those items' latent profiles, and thus both user-specific ratings and future popularity of the items can be predicted simultaneously. Specifically, such a joint prediction task is formulated as a two-stage optimization problem, and a sentinel user selection algorithm is devised to facilitate effective latent profiles extraction for both item ratings and popularity predictions. Experiments with microblogging and movie data sets corroborate that the proposed method is capable of mitigating the cold-start problem and it outperforms several competitive peer methods.
There is an increasing trend at using social media data to map human activity, to the point that some authors have suggested that these data may be even better than Earth Observation (EO) data to map urban areas. This work introduces a novel approach to map urban areas using SAR images from Sentinel-1 data, and exploiting either Twitter or Weibo data for two different cities, Beijing and Taipei. The use of different social media data results into different urban extent maps, and for different cities the same approach provides different accuracy values for the same area, according to the availability of either social media data set. Instead, urban extents obtained by means of EO data only are more stable. Indeed, the adoption of a specific social platform depends on a number of geographical, economic and political factors, while remote sensing is based on physics.
With the continuous development of AI, more and more people are experiencing "technology changes life". In artificial intelligence (AI) applications, intelligent conversation assistants play an important role, such as ChatGPT. Practitioners in the financial field have to deal with a huge amount of information every day, such as product price changes, market news, so they urgently need a reliable intelligent assistant to help them solve problems. We build a financial intelligent dialog assistant based on large language models (LLMs), which provides real-time retrieval, market analysis and many other functions. Multiple experiments prove the usability of our assistant.
Chinese financial derivatives market is both complicated and highly leveraged. Investors and regulators of derivatives market are required to aggregate information efficiently for detecting potential risks and keeping track of investment opportunities. Gathering information from financial news reports could be modeled as a multi-label text classification task. Many studies have been conducted in this area. However, most previous studies are designed to address common issues rather than specializing in financial area. These studies focus on the texts themselves while omitting supportive financial knowledge. Meanwhile, many previous studies employ a hierarchical label system, which may not be suitable for financial derivatives market because relations among financial labels are web-like rather than tree-like. To draw these limitations when aggregating news report information of Chinese financial derivatives market, this paper models texts to be classified and label structures simultaneously. A financial knowledge graph about Chinese financial derivatives market is firstly built. Then representative labels describing Chinese financial derivatives market are chosen. Necessary financial knowledge for classification task is finally extracted from the knowledge graph and merged into a multi-label text classification model. Experimental results on actual financial derivatives related data show that our method brings improvements compared to methods without label networks and methods without financial knowledge.