Book
WSDM '20: Proceedings of the 13th International Conference on Web Search and Data Mining
Proceedings of the 13th International Conference onWeb Search and Data Mining
We analyze comparative questions, i.e., questions asking to compare different items, that were submitted to Yandex in 2012. Responses to such questions might be quite different from the simple "ten blue links'' and could, for example, aggregate pros and cons of the different options as direct answers. However, changing the result presentation is an intricate decision such that the classification of comparative questions forms a highly precision-oriented task.
From a year-long Yandex log, we annotate a random sample of 50,000~questions; 2.8% of which are comparative. For these annotated questions, we develop a precision-oriented classifier by combining carefully hand-crafted lexico-syntactic rules with feature-based and neural approaches---achieving a recall of 0.6 at a perfect precision of 1.0. After running the classifier on the full year log (on average, there is at least one comparative question per second), we analyze 6,250 comparative questions using more fine-grained subclasses (e.g., should the answer be a "simple'' fact or rather a more verbose argument) for which individual classifiers are trained. An important insight is that more than 65% of the comparative questions demand argumentation and opinions, i.e., reliable direct answers to comparative questions require more than the facts from a search engine's knowledge graph.
In addition, we present a qualitative analysis of the underlying comparative information needs (separated into 14 categories likeconsumer electronics or health), their seasonal dynamics, and possible answers from community question answering platforms.
Recent research has shown the advantages of using autoencoders based on deep neural networks for collaborative filtering. In particular, the recently proposed Mult-VAE model, which used the multinomial likelihood variational autoencoders, has shown excellent results for top-N recommendations. In this work, we propose the Recommender VAE (RecVAE) model that originates from our research on regularization techniques for variational autoencoders. RecVAE introduces several novel ideas to improve Mult-VAE, including a novel composite prior distribution for the latent codes, a new approach to setting the β hyperparameter for the β-VAE framework, and a new approach to training based on alternating updates. In experimental evaluation, we show that RecVAE significantly outperforms previously proposed autoencoder-based models, including Mult-VAE and RaCT, across classical collaborative filtering datasets, and present a detailed ablation study to assess our new developments.
The rapid growth of geospatial data in the world enables the implementation of data mining techniques to mine the patterns in geospatial data. In this paper the authors have applied the algorithms that were previously used for mining slightly changing patterns in time series to geospatial data of the real estate market. So the paper discusses mining the patterns that slightly change in space (instead of time). The paper uses data on the real estate market. The predicted variable (square meter price) is analyzed respective to the district, distance to the city center, stations of public transport, highways, shops, sports, entertainment, healthcare, education centers, offices, parks etc. The proposed approach for mining slightly changing patterns in geospatial data is highly applicable to any data with geo-tag, e.g. space image recognition, geo-targeted marketing etc.
We analyze comparative questions, i.e., questions asking to compare different items, that were submitted to Yandex in 2012. Responses to such questions might be quite different from the simple "ten blue links'' and could, for example, aggregate pros and cons of the different options as direct answers. However, changing the result presentation is an intricate decision such that the classification of comparative questions forms a highly precision-oriented task.
From a year-long Yandex log, we annotate a random sample of 50,000~questions; 2.8% of which are comparative. For these annotated questions, we develop a precision-oriented classifier by combining carefully hand-crafted lexico-syntactic rules with feature-based and neural approaches---achieving a recall of 0.6 at a perfect precision of 1.0. After running the classifier on the full year log (on average, there is at least one comparative question per second), we analyze 6,250 comparative questions using more fine-grained subclasses (e.g., should the answer be a "simple'' fact or rather a more verbose argument) for which individual classifiers are trained. An important insight is that more than 65% of the comparative questions demand argumentation and opinions, i.e., reliable direct answers to comparative questions require more than the facts from a search engine's knowledge graph.
In addition, we present a qualitative analysis of the underlying comparative information needs (separated into 14 categories likeconsumer electronics or health), their seasonal dynamics, and possible answers from community question answering platforms.
This book constitutes the refereed proceedings of the 11th International Conference on Intelligent Data Processing, IDP 2016, held in Barcelona, Spain, in October 2016.
The 11 revised full papers were carefully reviewed and selected from 52 submissions. The papers of this volume are organized in topical sections on machine learning theory with applications; intelligent data processing in life and social sciences; morphological and technological approaches to image analysis.
The article discusses the influence of temperament on the academic performance of the first-year students at HSE-Nizhny Novgorod on the example of the Faculty of Informatics, Mathematics and Computer Science (IM&CS). The analyses were done with the help of statistics and educational data mining. The baseline data for the study is information about students, obtained by a survey: the information about temperament, degree of extraversion, stability, and other personality traits of students. The study involved students of the first and second years of the faculty of the IM&CS 2017-2018 academic year. Further, psychological factors affecting the average score and the probability of re-training for students with different temperaments were identified. A certain connection between temperament and academic success, which makes possible the prediction of "risky" students, was found. Various machine learning methods are used: the kNN-method and decision trees. The best results were shown by decision trees. As a result, first-year students are classified into three groups (Good, Medium, Bad) according to the degree of risk of getting academic debt. The practical result of the research was the recommendations to the educational office of the Faculty of IM&CS to pay attention to risky students and assist them in the educational process. After the end of the summer session, the classification results were checked. The article also presents an algorithm for finding risky students, taking temperament into account.