Recent advances in language model (LM) agents and function calling have enabled autonomous, feedback-driven systems to solve problems across various digital domains. To better understand the unique limitations of LM agents, we introduce RefactorBench, a benchmark consisting of 100 large handcrafted multi-file refactoring tasks in popular open-source repositories. Solving tasks within RefactorBench requires thorough exploration of dependencies across multiple files and strong adherence to relevant instructions. Every task is defined by 3 natural language instructions of varying specificity and is mutually exclusive, allowing for the creation of longer combined tasks on the same repository. Baselines on RefactorBench reveal that current LM agents struggle with simple compositional tasks, solving only 22\% of tasks with base instructions, in contrast to a human developer with short time constraints solving 87\%. Through trajectory analysis, we identify various unique failure modes of LM agents, and further explore the failure mode of tracking past actions. By adapting a baseline agent to condition on representations of state, we achieve a 43.9\% improvement in solving RefactorBench tasks. We further extend our state-aware approach to encompass entire digital environments and outline potential directions for future research. RefactorBench aims to support the study of LM agents by providing a set of real-world, multi-hop tasks within the realm of code.

Software version migration and program translation are an important and costly part of the lifecycle of large codebases.Traditional machine translation relies on parallel corpora for supervised translation, which is not feasible for program translation due to a dearth of aligned data.Recent unsupervised neural machine translation techniques have overcome data limitations by included techniques such as back translation and low level compiler intermediate representations (IR).These methods face significant challenges due to the noise in code snippet alignment and the diversity of IRs respectively.In this paper we propose a novel model called Code Distillation (CoDist) whereby we capture the semantic and structural equivalence of code in a language agnostic intermediate representation.Distilled code serves as a translation pivot for any programming language, leading by construction to parallel corpora which scale to all available source code by simply applying the distillation compiler.We demonstrate that our approach achieves state-of-the-art performance on CodeXGLUE and TransCoder GeeksForGeeks translation benchmarks, with an average absolute increase of 12.7% on the TransCoder GeeksforGeeks translation benchmark compare to TransCoder-ST.

Program translation, i.e. transcompilation has been attracting increasing attention from researchers due to its enormous application value.However, we observe that current program translating models still make elementary syntax errors, particularly when the source language uses syntax elements not present in the target language, which is exactly what developers are concerned about while may not be well exposed by frequently used metrics such as BLEU, CodeBLEU and Computation Accuracy.In this paper, we focus on evaluating the model's ability to address these basic syntax errors and developed an novel active defects probing suite, the Syntactic Unit Tests (SUT) and highly interpretable evaluation harness including Syntax Unit Test Accuracy (SUT Acc) metric and Syntax Element Test Score (SETS), to help diagnose and promote progress in this area.Our Syntactic Unit Test fills the gap in the community for a fine-grained evaluation dataset for program translation.Experimental analysis shows that our evaluation harness is more accurate, reliable, and in line with human judgments compared to previous metrics.

Software development life cycle is profoundly influenced by bugs; their introduction, identification, and eventual resolution account for a significant portion of software development cost. This has motivated software engineering researchers and practitioners to propose different approaches for automating the identification and repair of software defects.

In software development, it is common for programmers to copy-paste or port code snippets and then adapt them to their use case. This scenario motivates the code adaptation task – a variant of program repair which aims to adapt variable identifiers in a pasted snippet of code to the surrounding, preexisting context. However, no existing approach has been shown to effectively address this task. In this paper, we introduce AdaptivePaste, a learning-based approach to source code adaptation, based on transformers and a dedicated dataflow-aware deobfuscation pre-training task to learn meaningful representations of variable usage patterns. We demonstrate that AdaptivePaste can learn to adapt Python source code snippets with 67.8% exact match accuracy. We study the impact of confidence thresholds on the model predictions, showing the model precision can be further improved to 85.9% with our parallel-decoder transformer model in a selective code adaptation setting. To assess the practical use of AdaptivePaste we perform a user study among Python software developers on real-world copy-paste instances. The results show that AdaptivePaste reduces dwell time to nearly half the time it takes to port code manually, and helps to avoid bugs. In addition, we utilize the participant feedback to identify potential avenues for improvement.

Unit testing represents the foundational basis of the software testing pyramid, beneath integration and end-to-end testing. Automated software testing researchers have proposed a variety of techniques to assist developers in this time-consuming task.

Due to increasingly complex software design and rapid iterative development, code defects and security vulnerabilities are prevalent in modern software. In response, programmers rely on static analysis tools to regularly scan their codebases and find potential bugs. In order to maximize coverage, however, these tools generally tend to report a significant number of false positives, requiring developers to manually verify each warning. To address this problem, we propose a Transformer-based learning approach to identify false positive bug warnings. We demonstrate that our models can improve the precision of static analysis by 17.5%. In addition, we validated the generalizability of this approach across two major bug types: null dereference and resource leak.

Continuous evolution in modern software often causes documentation, tutorials, and examples to be out of sync with changing interfaces and frameworks. Relying on outdated documentation and examples can lead programs to fail or be less efficient or even less secure. In response, programmers need to regularly turn to other resources on the web, such as StackOverflow for examples to guide them in writing software. We recognize that this inconvenient, error-prone, and expensive process can be improved by using machine learning applied to software usage data. In this paper, we present a practical system, which uses machine learning on large-scale telemetry data and documentation corpora, generating appropriate and complex examples that can be used to improve documentation. We discuss both feature-based and transformer-based machine learning approaches and demonstrate that our system achieves 100% coverage for the used functionalities in the product, providing up-to-date examples upon every release and reduces the numbers of PRs submitted by software owners writing and editing documentation by >68%. We also share valuable lessons learnt during the 3 years that our production quality system has been deployed for Azure Cloud Command Line Interface (Azure CLI)

Unit testing is an essential part of the software development process, which helps to identify issues with source code in early stages of development and prevent regressions. Machine learning has emerged as viable approach to help software developers generate automated unit tests. However, generating reliable unit test cases that are semantically correct and capable of catching software bugs or unintended behavior via machine learning requires large, metadata-rich, datasets. In this paper we present Methods2Test: a large, supervised dataset of test cases mapped to corresponding methods under test (i.e., focal methods). This dataset contains 780,944 pairs of JUnit tests and focal methods, extracted from a total of 91,385 Java open source projects hosted on GitHub with licenses permitting re-distribution. The main challenge behind the creation of the Methods2Test was to establish a reliable mapping between a test case and the relevant focal method. To this aim, we designed a set of heuristics, based on developers' best practices in software testing, which identify the likely focal method for a given test case. To facilitate further analysis, we store a rich set of metadata for each method-test pair in JSON-formatted files. Additionally, we extract textual corpus from the dataset at different context levels, which we provide both in raw and tokenized forms, in order to enable researchers to train and evaluate machine learning models for Automated Test Generation. Methods2Test is publicly available at: https://github.com/microsoft/methods2test

Improving software performance is an important yet challenging part of the software development cycle. Today, the majority of performance inefficiencies are identified and patched by performance experts. Recent advancements in deep learning approaches and the wide-spread availability of open-source data creates a great opportunity to automate the identification and patching of performance problems. In this paper, we present DeepDev-PERF, a transformer-based approach to suggest performance improvements for C# applications. We pretrain DeepDev-PERF on English and Source code corpora, followed by finetuning for the task of generating performance improvement patches for C# applications. Our evaluation shows that our model can generate the same performance improvement suggestion as the developer fix in ‍53

Code review is an essential part to software development lifecycle since it aims at guaranteeing the quality of codes. Modern code review activities necessitate developers viewing, understanding and even running the programs to assess logic, functionality, latency, style and other factors. It turns out that developers have to spend far too much time reviewing the code of their peers. Accordingly, it is in significant demand to automate the code review process. In this research, we focus on utilizing pre-training techniques for the tasks in the code review scenario. We collect a large-scale dataset of real-world code changes and code reviews from open-source projects in nine of the most popular programming languages. To better understand code diffs and reviews, we propose CodeReviewer, a pre-trained model that utilizes four pre-training tasks tailored specifically for the code review scenario. To evaluate our model, we focus on three key tasks related to code review activities, including code change quality estimation, review comment generation and code refinement. Furthermore, we establish a high-quality benchmark dataset based on our collected data for these three tasks and conduct comprehensive experiments on it. The experimental results demonstrate that our model outperforms the previous state-of-the-art pre-training approaches in all tasks. Further analysis show that our proposed pre-training tasks and the multilingual pre-training dataset benefit the model on the understanding of code changes and reviews.

Collaborative software development is an integral part of the modern software development life cycle, essential to the success of large-scale software projects. When multiple developers make concurrent changes around the same lines of code, a merge conflict may occur. Such conflicts stall pull requests and continuous integration pipelines for hours to several days, seriously hurting developer productivity. To address this problem, we introduce MergeBERT, a novel neural program merge framework based on token-level three-way differencing and a transformer encoder model. By exploiting the restricted nature of merge conflict resolutions, we reformulate the task of generating the resolution sequence as a classification task over a set of primitive merge patterns extracted from real-world merge commit data. Our model achieves 63–68% accuracy for merge resolution synthesis, yielding nearly a 3× performance improvement over existing semi-structured, and 2× improvement over neural program merge tools. Finally, we demonstrate that MergeBERT is sufficiently flexible to work with source code files in Java, JavaScript, TypeScript, and C# programming languages. To measure the practical use of MergeBERT, we conduct a user study to evaluate MergeBERT suggestions with 25 developers from large OSS projects on 122 real-world conflicts they encountered. Results suggest that in practice, MergeBERT resolutions would be accepted at a higher rate than estimated by automatic metrics for precision and accuracy. Additionally, we use participant feedback to identify future avenues for improvement of MergeBERT.

Large Transformer models achieved the state-of-the-art status for Natural Language Understanding tasks and are increasingly becoming the baseline model architecture for modeling source code. Transformers are usually pre-trained on large unsupervised corpora, learning token representations and transformations relevant to modeling generally available text, and are then fine-tuned on a particular downstream task of interest. While fine-tuning is a tried-and-true method for adapting a model to a new domain -- for example, question-answering on a given topic -- generalization remains an on-going challenge. In this paper, we explore and evaluate transformer model fine-tuning for personalization. In the context of generating unit tests for Java methods, we evaluate learning to personalize to a specific software project using several personalization techniques. We consider three key approaches: (i) custom fine-tuning, which allows all the model parameters to be tuned; (ii) lightweight fine-tuning, which freezes most of the model's parameters, allowing tuning of the token embeddings and softmax layer only or the final layer alone; (iii) prefix tuning, which keeps model parameters frozen, but optimizes a small project-specific prefix vector. Each of these techniques offers a trade-off in total compute cost and predictive performance, which we evaluate by code and task-specific metrics, training time, and total computational operations. We compare these fine-tuning strategies for code generation and discuss the potential generalization and cost benefits of each in various deployment scenarios.

Colin Clement, Shuai Lu, Xiaoyu Liu, Michele Tufano, Dawn Drain, Nan Duan, Neel Sundaresan, Alexey Svyatkovskiy. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021.

Simultaneously modeling source code and natural language has many exciting applications in automated software development and understanding.Pursuant to achieving such technology, we introduce PYMT5, the PYTHON method text-to-text transfer transformer, which is trained to translate between all pairs of PYTHON method feature combinations: a single model that can both predict whole methods from natural language documentation strings (docstrings) and summarize code into docstrings of any common style.We present an analysis and modeling effort of a large-scale parallel corpus of 26 million PYTHON methods and 7.7 million method-docstring pairs, demonstrating that for docstring and method generation, PYMT5 outperforms similarlysized auto-regressive language models (GPT2) which were English pre-trained or randomly initialized.On the CODE-SEARCHNET test set, our best model predicts 92.1% syntactically correct method bodies, achieved a BLEU score of 8.59 for method generation and 16.3 for docstring * Corresponding author † Work done during a Microsoft internship generation (summarization), and achieved a ROUGE-L F-score of 24.8 for method generation and 36.7 for docstring generation.

In software development through integrated development environments (IDEs), code completion is one of the most widely used features. Nevertheless, majority of integrated development environments only support completion of methods and APIs, or arguments.

The last decade has seen three great phenomena in computing - the rebirth of AI algorithms and AI hardware; the evolution of cloud computing and distributed software development; and the explosive growth of open source software that has led to the availability of code as data, and its associated metadata, at scale. In this talk, we will describe how we take advantage of innovations in these dimensions to improve developer productivity and infuse AI and automation into software processes. We will discuss examples of how we built intelligent software by creating AI algorithms driven by deep understanding of code as data. In addition, we will talk about how data can also be treated as code for the next-generation AI-infused software development.

In this paper, we propose a novel end-to-end approach for AI-assisted code completion called Pythia. It generates ranked lists of method and API recommendations which can be used by software developers at edit time. The system is currently deployed as part of Intellicode extension in Visual Studio Code IDE. Pythia exploits state-of-the-art large-scale deep learning models trained on code contexts extracted from abstract syntax trees. It is designed to work at a high throughput predicting the best matching code completions on the order of 100 ms.

We identify a novel instance of the background subtraction problem that focuses on extracting near-field foreground objects captured using handheld cameras. Given two user-generated videos of a scene, one with and the other without the foreground object (s), our goal is to efficiently generate an output video with only the foreground object (s) present in it. We cast this challenge as a spatio-temporal frame matching problem, and propose an efficient solution for it that exploits the temporal smoothness of the video sequences. We present theoretical analyses for the error bounds of our approach, and validate our findings using a detailed set of simulation experiments. Finally, we present the results of our approach tested on multiple real videos captured using handheld cameras, and compare them to several alternate foreground extraction approaches.

We describe a completely automated large scale visual recommendation system for fashion. Our focus is to efficiently harness the availability of large quantities of online fashion images and their rich meta-data. Specifically, we propose two classes of data driven models in the Deterministic Fashion Recommenders (DFR) and Stochastic Fashion Recommenders (SFR) for solving this problem. We analyze relative merits and pitfalls of these algorithms through extensive experimentation on a large-scale data set and baseline them against existing ideas from color science. We also illustrate key fashion insights learned through these experiments and show how they can be employed to design better recommendation systems. The industrial applicability of proposed models is in the context of mobile fashion shopping. Finally, we also outline a large-scale annotated data set of fashion images Fashion-136K) that can be exploited for future research in data driven visual fashion.

The rapid increase in number of high quality mobile cameras have opened up an array of new problems in mobile vision. Mobile cameras are predominantly monocular and are devoid of any sense of depth, making them heavily reliant on 2D image processing. Understanding 3D structure of scenes being imaged can greatly improve the performance of existing vision/graphics techniques. In this regard, recent availability of large scale RGB-D datasets beg for more effective data driven strategies to leverage the scale of data. We propose a depth recovery mechanism “im2depth”, that is lightweight enough to run on mobile platforms, while leveraging the large scale nature of modern RGB-D datasets. Our key observation is to form a basis (dictionary) over the RGB and depth spaces, and represent depth maps by a sparse linear combination of weights over dictionary elements. Subsequently, a prediction function is estimated between weight vectors in RGB to depth space to recover depth maps from query images. A final superpixel post processor aligns depth maps with occlusion boundaries, creating physically plausible results. We conclude with thorough experimentation with four state of the art depth recovery algorithms, and observe an improvement of over 6.5 percent in shape recovery, and over 10cm reduction in average L1 error.

In online peer-to-peer commerce places where physical examination of the goods is infeasible, textual descriptions, images of the products, reputation of the participants, play key roles. Visual image is a powerful channel to convey crucial information towards e-shoppers and influence their choice. In this paper, we investigate a well-known online marketplace where over millions of products change hands and most are described with the help of one or more images. We present a systematic data mining and knowledge discovery approach that aims to quantitatively dissect the role of images in e-commerce in great detail. Our goal is two-fold. First, we aim to get a thorough understanding of impact of images across various dimensions: product categories, user segments, conversion rate. We present quantitative evaluation of the influence of images and show how to leverage different image aspects, such as quantity and quality, to effectively raise sale. Second, we study interaction of image data with other selling dimensions by jointly modeling them with user behavior data. Results suggest that "watch" behavior encodes complex signals combining both attention and hesitation from buyer, in which image still holds an important role when compared to other selling variables, especially for products for which appearance is important. We conclude on how these findings can benefit sellers in a high competitive online e-commerce market.

The popularity of social media sites like Twitter and Facebook opens up interesting research opportunities for understanding the interplay of social media and e-commerce. Most research on online behavior, up until recently, has focused mostly on social media behaviors and e-commerce behaviors independently. In our study we choose a particular global e-commerce platform (eBay) and a particular global social media platform (Twitter). We quantify the characteristics of the two individual trends as well as the correlations between them. We provide evidences that about 5% of general eBay query streams show strong positive correlations with the corresponding Twitter mention streams, while the percentage jumps to around 25% for trending eBay query streams. Some categories of eBay queries, such as 'Video Games' and 'Sports', are more likely to have strong correlations. We also discover that eBay trend lags Twitter for correlated pairs and the lag differs across categories. We show evidences that celebrities' popularities on Twitter correlate well with their relevant search and sales on eBay. The correlations and lags provide predictive insights for future applications that might lead to instant merchandising opportunities for both sellers and e-commerce platforms.

Given the enormous growth in user-generated videos, it is becoming increasingly important to be able to navigate them efficiently. As these videos are generally of poor quality, summarization methods designed for well-produced videos do not generalize to them. To address this challenge, we propose to use web-images as a prior to facilitate summarization of user-generated videos. Our main intuition is that people tend to take pictures of objects to capture them in a maximally informative way. Such images could therefore be used as prior information to summarize videos containing a similar set of objects. In this work, we apply our novel insight to develop a summarization algorithm that uses the web-image based prior information in an unsupervised manner. Moreover, to automatically evaluate summarization algorithms on a large scale, we propose a framework that relies on multiple summaries obtained through crowdsourcing. We demonstrate the effectiveness of our evaluation framework by comparing its performance to that of multiple human evaluators. Finally, we present results for our framework tested on hundreds of user-generated videos.

With the explosion of mobile devices with cameras, online search has moved beyond text to other modalities like images, voice, and writing. For many applications like Fashion, image-based search offers a compelling interface as compared to text forms by better capturing the visual attributes. In this paper we present a simple and fast search algorithm that uses color as the main feature for building visual search. We show that low level cues such as color can be used to quantify image similarity and also to discriminate among products with different visual appearances. We demonstrate the effectiveness of our approach through a mobile shopping application\footnote{eBay Fashion App available at https://itunes.apple.com/us/app/ebay-fashion/id378358380?mt=8 and eBay image swatch is the feature indexing millions of real world fashion images}. Our approach outperforms several other state-of-the-art image retrieval algorithms for large scale image data.

Reuse and remarketing of content and products is an integral part of the internet. As E-commerce has grown, online resale and secondary markets form a significant part of the commerce space. The intentions and methods for reselling are diverse. In this paper, we study an instance of such markets that affords interesting data at large scale for mining purposes to understand the properties and patterns of this online market.As part of knowledge discovery of such a market, we first formally propose criteria to reveal unseen resale behaviors by elastic matching identification (EMI) based on the account transfer and item similarity properties of transactions. Then, we present a large-scale system that leverages MapReduce paradigm to mine millions of online resale activities from petabyte scale heterogeneous e-commerce data. With the collected data, we show that the number of resale activities leads to a power law distribution with a 'long tail', where a significant share of users only resell in very low numbers and a large portion of resales come from a small number of highly active resellers. We further conduct a comprehensive empirical study from different aspects of resales, including the temporal, spatial patterns, user demographics, reputation and the content of sale postings. Based on these observations, we explore the features related to "successful" resale transactions and evaluate if they can be predictable. We also discuss uses of this information mining for business insights and user experience on a real-world online marketplace.

In this paper we study diversity and its relations to search relevance in the context of an online marketplace. We conduct a large-scale log-based study using click-stream data from a leading eCommerce site. We introduce 3 main metrics -- selection (diversity), trust, and value. In our analysis we also show how these interact with relevance in different ways. We study the benefits of diversity and also show why guaranteeing diversity is important.

In today's world, brand based shopping is popular especially in product lines like clothing and shoes, appliances, and electronics. Because of the importance of brands while shopping, it has become important for online shopping portals to consider brand loyalty and brand preferences of users. In this paper, we describe a system designed for brand-based shopping and exploration. The system is built by analyzing a large query set consisting of 115M queries from eBay.com -- a vibrant marketplace with more than 95M active users. The system allows brand-pivoted exploration of inventory. It allows exploration and purchase of substitute branded goods (e.g. Sony camcorder for Canon camcorder) and complementary branded merchandise (e.g. Lego castle set for Lego train station set).

Hierarchical classification is a challenging problem yet bears a broad application in real-world tasks. Item categorization in the ecommerce domain is such an example. In a large-scale industrial setting such as eBay, a vast amount of items need to be categorized into a large number of leaf categories, on top of which a complex topic hierarchy is defined. Other than the scale challenges, item data is extremely sparse and skewed distributed over categories, and exhibits heterogeneous characteristics across categories. A common strategy for hierarchical classification is the "gates-and-experts" methods, where a high-level classification is made first (the gates), followed by a low-level distinction (the experts). In this paper, we propose to leverage domain-specific feature generation and modeling techniques to greatly enhance the classification accuracy of the experts. In particular, we innovatively derive features to encode various rich domain knowledge and linguistic hints, and then adapt a SVM-based model to distinguish several very confusing category groups appeared as the performance bottleneck of a currently deployed live system at eBay. We use illustrative examples and empirical results to demonstrate the effectiveness of our approach, particularly the merit of smartly designed domain-specific features.

Recommender systems form the core of e-commerce systems. In this paper we take a top-down view of recommender systems and identify challenges, opportunities, and approaches in building recommender systems for a marketplace platform. We use eBay as an example where the elaborate interaction offers a number opportunities for creative recommendations. However, eBay also poses complexities resulting from high sparsity of relationships. Our discussion can be generalized beyond eBay to other marketplaces.

In this paper, we design a recommender system for the post-purchase stage, i.e., after a user purchases a product. Our method combines both behavioral and content aspects of recommendations. We first find the most related categories for the active product in the post-purchase stage. Among these related categories, products with high behavioral relevance and content relevance are recommended to the user. In addition, our algorithm considers the temporal factor, i.e., the purchase time of the active product and the recommendation time. We apply our algorithm on a random sample of the purchase data from eBay. Comparing to the baseline item-based collaborative filtering approach, our hybrid recommender system achieves significant coverage and purchase rate gain for different time windows.

User expectation and experience for web search and eCommerce (product) search are quite different. Product descriptions are concise as compared to typical web documents. User expectation is more specific to find the right product. The difference in the publisher and searcher vocabulary (in case of product search the seller and the buyer vocabulary) combined with the fact that there are fewer products to search over than web documents result in observable numbers of searches that return no results (zero recall searches). In this paper we describe a study of zero recall searches. Our study is focused on eCommerce search and uses data from a leading eCommerce site's user click stream logs. There are 3 main contributions of our study: 1) The cause of zero recall searches; 2) A study of user's reaction and recovery from zero recall; 3) A study of differences in behavior of power users versus novice users to zero recall searches.

Query suggestion module is an integral part of every search engine. It helps search engine users narrow or broaden their searches. Published work on query suggestion methods has mainly focused on the web domain. But, the module is also popular in the domain of e-commerce for product search. In this paper, we discuss query suggestion and its methodologies in the context of e-commerce search engines. We show that dynamic inventory combined with long and sparse tail of query distribution poses unique challenges to build a query suggestion method for an e-commerce marketplace. We compare and contrast the design of a query suggestion system for web search engines and e-commerce search engines. Further, we discuss interesting measures to quantify the effectiveness of our query suggestion methodologies. We also describe the learning gained from exposing our query suggestion module to a vibrant community of millions of users.

Commerce networks involve buying and selling activities among individuals or organizations. As the growing of the Internet and e-commerce, it brings opportunities for obtaining real world online commerce networks, which are magnitude larger than before. Getting a deeper understanding of e-commerce networks, such as the eBay marketplace, in terms of what structure they have, what kind of interactions they afford, what trust and reputation measures exist, and how they evolve has tremendous value in suggesting business opportunities and building effective user applications. In this paper, we modeled the eBay network as a complex network. We analyzed the macroscopic shape of the network using degree distribution and the bow-tie model. Networks of different eBay categories are also compared. The results suggest that the categories vary from collector networks to retail networks. We also studied the local structures of the networks using motif profiling. Finally, patterns of preferential connections are visually analyzed using Auroral diagrams.

One central issue in a long-tail online marketplace such as eBay is to automatically put user self-input items into a catalog in real time. This task is extremely challenging when the inventory scales up, the items become ephemeral, and the user input remains noisy. Indeed, catalog learning has emerged as a key technical property for other major online ecommerce applications including search and recommendation. We formulate the item cataloging task as a Bayesian classification problem, which shall scale well in very large data set and have good online prediction performance. The inherent data sparseness issue, especially for those tail categories, is key to the overall model performance. We address the data sparseness issue by adapting statistically sound smoothing methods well studied in language modeling tasks. However, there are data characteristics specific to the ecommerce domain, including short yet focused item description, very large and hierarchical catalog taxonomy, and highly skewed distribution over types of items. We investigate these domain-specific regularities empirically, and report practically significant results with real-world true-scale data.

Web 2.0 technologies have enabled more and more people to freely comment on different kinds of entities (e.g. sellers, products, services). The large scale of information poses the need and challenge of automatic summarization. In many cases, each of the user-generated short comments comes with an overall rating. In this paper, we study the problem of generating a ``rated aspect summary'' of short comments, which is a decomposed view of the overall ratings for the major aspects so that a user could gain different perspectives towards the target entity. We formally define the problem and decompose the solution into three steps. We demonstrate the effectiveness of our methods by using eBay sellers' feedback comments. We also quantitatively evaluate each step of our methods and study how well human agree on such a summarization task. The proposed methods are quite general and can be used to generate rated aspect summary automatically given any collection of short comments each associated with an overall rating.

In this paper, we describe a buzz-based recommender system based on a large source of queries in an eCommerce application. The system detects bursts in query trends. These bursts are linked to external entities like news and inventory information to find the queries currently in-demand which we refer to as buzz queries. The system follows the paradigm of limited quantity merchandising, in the sense that on a per-day basis the system shows recommendations around a single buzz query with the intent of increasing user curiosity, and improving activity and stickiness on the site. A semantic neighborhood of the chosen buzz query is selected and appropriate recommendations are made on products that relate to this neighborhood.

In this paper we describe how high quality transaction data comprising of online searching, product viewing, and product buying activity of a large online community can be used to infer semantic relationships between queries. We work with a large scale query log consisting of around 115 million queries from eBay. We discuss various techniques to infer semantic relationships among queries and show how the results from these methods can be combined to measure the strength and depict the kinds of relationships. Further, we show how this extraction of relations can be used to improve search relevance, related query recommendations, and recovery from null results in an eCommerce context.

In this paper, we present an outline of a software system for buzz-based recommendations. This system is based on a large source of queries in an eCommerce application. The buzz events are detected based on query bursts linked to external entities like news and inventory information. A semantic neighborhood of the chosen buzz query is selected and appropriate recommendations are made on products that relate to this neighborhood. The system follows the paradigm of limited quantity merchandizing, in the sense that on a per-day basis the system shows recommendations around a single buzz query with the intent of increasing user curiosity and promoting user activity and stickiness. The system demonstrates the deployment of an interesting application based on KDD principles applied to a high volume industrial context.

In large scale online systems like Search, eCommerce, or social network applications, user queries represent an important dimension of activities that can be used to study the impact on the system, and even the business. In this paper, we describe how to detect, characterize and classify bursts in user queries in a large scale eCommerce system. We build upon the approaches discussed in KDD 2002 "Bursty and Hierarchical Structure in Streams" [3] and apply them to a high volume industrial context. We describe how to identify bursts on a near real-time basis, classify them, and apply them to build interesting merchandizing applications.

In this paper we describe our mining system which automatically mines tags from feedback text in an eCommerce scenario. It renders these tags in a visually appealing manner. Further, emoticons are attached to mined tags to add sentiment to the visual aspect.

As online commerce, social networks, and user-generated content become common, need for trust and reputation models become prime. This tutorial will give an overview of trust and reputation systems as studied by social network researchers. Other topics include: Reputation and its relationship to security and fraud; Feedback and other Manifestations and Implementations of trust and reputations; Models of reputation; platforms: P2P systems, Centralized systems; Auctions, Incentive systems; Collaborative filtering; Social Networking and Social reputation; Portability and Universality of Identity, trust, and reputation.

Search engines provide a small window to the vast repository of data they index and against which they search. They try their best to return the documents that are of relevance to the user but often a large number of results may be returned. Users struggle to manage this vast result set looking for the items of interest. Clustering search results is one way of alleviating this navigational pain. In this paper we describe a clustering system that enables clustering search results in an online marketplace search system.

Despite the advancement of XML, the majority of documents on the Web is still marked up with HTML for visual rendering purposes only, thus building a huge amount of legacy data. In order to facilitate querying Web based data in a way more efficient and effective than just keyword based retrieval, enriching such Web documents with both structure and semantics is necessary. We describe a novel approach to the integration of topic specific HTML documents into a repository of XML documents. In particular, we describe how topic specific HTML documents are transformed into XML documents. The proposed document transformation and semantic element tagging process utilizes document restructuring rules and minimum information about the topic in the form of concepts. For the resulting XML documents, a majority schema is derived that describes common structures among the documents in the form of a DTD. We explore and discuss different techniques, and rules for document conversion and majority schema discovery. We finally demonstrate the feasibility and effectiveness of our approach by applying it to a set of resume HTML documents gathered by a Web crawler.

Article Share on Algorithms and programming models for efficient representation of XML for Internet applications Authors: Neel Sundaresan NehaNet Corp., 2355 Paragon Drive Suite F, San Jose, CA NehaNet Corp., 2355 Paragon Drive Suite F, San Jose, CAView Profile , Reshad Moussa NehaNet Corp., 2355 Paragon Drive Suite F, San Jose, CA NehaNet Corp., 2355 Paragon Drive Suite F, San Jose, CAView Profile Authors Info & Claims WWW '01: Proceedings of the 10th international conference on World Wide WebMay 2001Pages 366–375https://doi.org/10.1145/371920.372090Published:01 April 2001Publication History 18citation1,170DownloadsMetricsTotal Citations18Total Downloads1,170Last 12 Months1Last 6 weeks0 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Among the most critical issues of the internet today is how to make Web content accessible to all users, especially to users with disabilities. To meet the diverse needs and abilities of this population, the Web today calls for the development of new systems and methods to enable the same content to be adapted for display according to specific, often conflicting needs. One way to achieve this goal is through Web con- tent transcoding. This paper presents a system, called Aurora, that transcodes Web content based on semantic rather than syntactic constructs. The goal is to deliver Web-based services (such as auction, search engine, travel, etc.) to a diverse set of users according to their specific needs. Using a schema-driven framework, Aurora extracts and maps Web content into domain-specific XML data based on abstract user goals. In doing so, it separates the meaning Web content from its presentation. The system further enables an extensible set of interface adaptors to generate custom Web pages, on-the-fly, from this standardized XML data. Ultimately, it streamlines and customizes the Web interface to faciliate navigation. The mechanisms of this rule-based semantic transcoding system and its advantages and limitations as a strategy to make Web services more accessible are the subject of this paper.

This paper presents a relevant term discoverer, a system that discovers relevant topics of a given topic from the World Wide Web. The system mines hyperlink metadata on the basis of the association of terms in the metadata. It also applies various filtering techniques to detect false positives and false negatives. The applications of the system include: i) topic-specific information gathering systems that need to crawl resources of the relevant topic, ii) bibliography search system that need to extend their search to the articles of relevant topics, iii) classification systems that can categorize items of similar class together, and so on. We report a successful application of the system to build a topic-specific search-engine dedicated to eXtensible Markup Language (XML). Using the algorithms presented in this paper, we were able to identify the relevant topics that the search engine needs to cover. Together with effective topic-directed crawling algorithms, we were able to build a topic-specific search engine that require significantly less human labor but perform almost as well as topic-specific search engines whose content is maintained by humans.

In this paper, w e describe a novel text classi er that can eectiv ely cope with structured documents.We r e p o r t e xperiments that compare its performance with that of a wellknown probabilistic classi er.Our novel classi er can take adv antage of the information in the structure of document that con ven tional, purely term-based classi ers ignore.Conven tional classi ers are mostly based on the vector space model of document, which views a document simply as an n-dimensional vector of terms.T o retain the information in the structure, w e ha ve dev eloped a structured vector model, which represents a document with a structured vector, whose elements can be either terms or other structured vectors.With this extended model, we a l s o h a ve i m p r o ved the well-kno wn probabilistic classi cation method based on the Bernoulli document generation model.Our classi er based on these improvements performes signi cantly better on pre-classi ed samples from the web and the US Patent database than the usual classi ers.