Browsing by Author "Wang, Xinyue"

Now showing 1 - 6 of 6
  • Thumbnail Image
    Big Data Text Summarization: Using Deep Learning to Summarize Theses and Dissertations
    Ahuja, Naman; Bansal, Ritesh; Ingram, William A.; Jude, Palakh; Kahu, Sampanna; Wang, Xinyue (Virginia Tech, 2018-12-05)
    Team 16 in the fall 2018 course "CS 4984/5984 Big Data Text Summarization," in partnership with the University Libraries and the Digital Library Research Laboratory, prepared a corpus of electronic theses and dissertations (ETDs) for students to study natural language processing with the power of state-of-the-art deep learning technology. The ETD corpus is made up of 13,071 doctoral dissertations and 17,890 master theses downloaded from the University Libraries’ VTechWorks system. This particular study is designed to explore big data summarization for ETDs, which is a relatively under-explored area. The result of the project will help to address the difficulty of information extraction from ETD documents, the potential of transfer learning on automatic summarization of ETD chapters, and the quality of state-of-the-art deep learning summarization technologies when applied to the ETD corpus. The goal of this project is to generate chapter level abstractive summaries for an ETD collection through deep learning. Major challenges of the project include accurately extracting well-formatted chapter text from PDF files, and the lack of labeled data for supervised deep learning models. For PDF processing, we compare two state of the art scholarly PDF data extraction tools, Grobid and Science-Parse, which generate structured documents from which we can further extract metadata and chapter level text. For the second challenge, we perform transfer learning by training supervised learning models on a labeled dataset of Wikipedia articles related to the ETD collection. Our experimental models include Sequence-to-Sequence and Pointer Generator summarization models. Besides supervised models, we also experiment with an unsupervised reinforcement model, Fast Abstractive Summarization-RL. The general pipeline for our experiments consists of the following steps: PDF data processing and chapter extraction, collecting a training data set of Wikipedia articles, manually creating human generated gold standard summaries for testing and validation, building deep learning models for chapter summarization, evaluating and tuning the models based on results, and then iteratively refining the whole process.
  • Thumbnail Image
    The Case For Alternative Web Archival Formats To Expedite The Data-To-Insight Cycle
    Wang, Xinyue; Xie, Zhiwu (ACM, 2020-08)
    The WARC file format is widely used by web archives to preserve collected web content for future use. With the rapid growth of web archives and the increasing interest to reuse these archives as big data sources for statistical and analytical research, the speed to turn these data into insights becomes critical. In this paper we show that the WARC format carries significant performance penalties for batch processing workload. We trace the root cause of these penalties to its data structure, encoding, and addressing method. We then run controlled experiments to illustrate how severe these problems can be. Indeed, performance gain of one to two orders of magnitude can be achieved simply by reformatting WARC files into Parquet or Avro formats. While these results do not necessarily constitute an endorsement for Avro or Parquet, the time has come for the web archiving community to consider replacing WARC with more efficient web archival formats.
  • Thumbnail Image
    CS5604 Information Storage and Retrieval Fall 2017 Solr Report
    Kumar, Abhinav; Bangad, Anand; Robertson, Jeff; Garg, Mohit; Ramesh, Shreyas; Mi, Siyu; Wang, Xinyue; Wang, Yu (Virginia Tech, 2018-01-15)
    The Digital Library Research Laboratory (DLRL) has collected over 1.5 billion tweets and millions of webpages for the Integrated Digital Event Archiving and Library (IDEAL) and Global Event Trend Archive Research (GETAR) projects. We are using a 21 node Cloudera Hadoop cluster to store and retrieve this information. One goal of this project is to expand the data collection to include more web archives and geospatial data beyond what previously had been collected. Another important part in this project is optimizing the current system to analyze and allow access to the new data. To accomplish these goals, this project is separated into 6 parts with corresponding teams: Classification (CLA), Collection Management Tweets (CMT), Collection Management Webpages (CMW), Clustering and Topic Analysis (CTA), Front-end (FE), and SOLR. The report describes the work completed by the SOLR team which improves the current searching and storage system. We include the general architecture and an overview of the current system. We present the part that Solr plays within the whole system with more detail. We talk about our goals, procedures, and conclusions on the improvements we made to the current Solr system. This report also describes how we coordinate with other teams to accomplish the project at a higher level. Additionally, we provide manuals for future readers who might need to replicate our experiments. The main components within the Cloudera Hadoop cluster that the SOLR team interacts with include: Solr searching engine, HBase database, Lily indexer, Hive database, HDFS file system, Solr recommendation plugin, and Mahout. Our work focuses on HBase design, data quality control, search recommendations, and result ranking. Overall, throughout the semester, we have processed 12,564 web pages and 5.9 million tweets. In order to cooperate with Geo Blacklight, we make major changes on the Solr schema. We also function as a data quality control gateway for the Front End team and deliver the finalized data for them. As to search recommendation, we provide search recommendation such as the MoreLikeThis plugin within Solr for recommending related records from search results, and a custom recommendation system based on user behavior to provide user based search recommendations. After the fine tuning over the final weeks of semester, we successfully allowed effective connection of results from data provided by other teams, and delivered them to the front end through a Solr core.
  • Thumbnail Image
    Large Web Archive Collection Infrastructure and Services
    Wang, Xinyue (Virginia Tech, 2023-01-20)
    The web has evolved to be the primary carrier of human knowledge during the information age. The ephemeral nature of much web content makes web knowledge preservation vital in preserving human knowledge and memories. Web archives are created to preserve the current web and make it available for future reuse. A growing number of web archive initia- tives are actively engaging in web archiving activities. Web archiving standards like WARC, for formatted storage, have been established to standardize the preservation of web archive data. In addition to its preservation purpose, web archive data is also used as a source for research and for lost information recovery. However, the reuse of web archive data is inherently challenging because of the scale of data size and requirements of big data tools to serve and analyze web archive data efficiently. In this research, we propose to build web archive infrastructure that can support efficient and scalable web archive reuse with big data formats like Parquet, enabling more efficient quantitative data analysis and browsing services. Upon the Hadoop big data processing platform with components like Apache Spark and HBase, we propose to replace the WARC (web archive) data format with a columnar data format Parquet to facilitate more efficient reuse. Such a columnar data format can provide the same features as WARC for long-term preservation. In addition, the columnar data format introduces the potential for better com- putational efficiency and data reuse flexibility. The experiments show that this proposed design can significantly improve quantitative data analysis tasks for common web archive data usage. This design can also serve web archive data for a web browsing service. Unlike the conventional web hosting design for large data, this design primarily works on top of the raw large data in file systems to provide a hybrid environment around web archive reuse. In addition to the standard web archive data, we also integrate Twitter data into our design as part of web archive resources. Twitter is a prominent source of data for researchers in a vari- ety of fields and an integral element of the web's history. However, Twitter data is typically collected through non-standardized tools for different collections. We aggregate the Twitter data from different sources and integrate it into the suggested design for reuse. We are able to greatly increase the processing performance of workloads around social media data by overcoming the data loading bottleneck with a web-archive-like Parquet data format.
  • Thumbnail Image
    Numerically Trained Ultrasound AI for Monitoring Tool Degradation
    Jin, Yuqi; Wang, Xinyue; Fox, Edward A.; Xie, Zhiwu; Neogi, Arup; Mishra, Rajiv S.; Wang, Tianhao (Wiley, 2022-01-13)
    Monitoring tool degradation during manufacturing can ensure product accuracy and reliability. However, due to variations in degradation conditions and complexity in signal analysis, effective and broadly applicable monitoring is still challenging to achieve. Herein, a novel monitoring method using ultrasound signals augmented with a numerically trained machine learning technique is reported to monitor the wear condition of friction stir welding and processing tools. Ultrasonic signals travel axially inside the tools, and even minor tool wear will change the time and amplitude of the reflected signal. An artificial intelligence (AI) algorithm is selected as a suitable referee to identify the small variations in the tool conditions based on the reflected ultrasound signals. To properly train the AI referee, a human‐error‐free data bank using numerical simulation is generated. The simulation models the experimental conditions with high fidelity and can provide comparable ultrasound signals. As a result, the trained AI model can recognize the tool wear from real experiments with subwavelength accuracy prediction of the worn amount on the tool pins.
  • Thumbnail Image
    Toward an Intelligent Crawling Scheduler for Archiving News Websites Using Reinforcement Learning
    Wang, Xinyue; Ahuja, Naman; Llorens, Nathaniel; Bansal, Ritesh; Dhar, Siddharth (Virginia Tech, 2019-12-03)
    Web crawling is one of the fundamental activities for many kinds of web technology organizations and companies such as Internet Archive and Google. While companies like Google often focus on content delivery for users, web archiving organizations such as the Internet Archive pay more attention to the accurate preservation of the web. Crawling accuracy and efficiency are major concerns in this task. An ideal crawling module should be able to keep up with the changes in the target web site with minimal crawling frequency to maximize the routine crawling efficiency. In this project, we investigate using information from web archives' history to help the crawling process within the scope of news websites. We aim to build a smart crawling module that can predict web content change accurately both on the web page and web site structure level through modern machine learning algorithms and deep learning architectures. At the end of the project: We have collected and processed raw web archive collections from Archive.org and through our frequent crawling jobs. We have developed methods to extract identical copies of web page content and web site structure from the web archive data. We have implemented baseline models for predicting web page content change and web site structure change, web page content change with supervised machine learning algorithms; We have implemented two different reinforcement learning models for generating a web page crawling plan: a continuous prediction model and a sparse prediction model. Our results show that the reinforcement learning modal has the potential to work as an intelligent web crawling scheduler.