Message Passing Neural Networks (MPNNs) hold a key position in machine learning on graphs, but they struggle with unintended behaviors, such as over-smoothing and over-squashing, due to irregular data structures. The observation and formulation of these limitations have become foundational in constructing more informative graph representations. In this paper, we delve into the limitations of MPNNs, focusing on aspects that have previously been overlooked. Our observations reveal that even within a single layer, the information specific to an individual node can become significantly diluted. To delve into this phenomenon in depth, we present the concept of Over-dilution and formulate it with two dilution factors: intra-node dilution for attribute-level and inter-node dilution for node-level representations. We also introduce a transformer-based solution that alleviates over-dilution and complements existing node embedding methods like MPNNs. Our findings provide new insights and contribute to the development of informative representations. The implementation and supplementary materials are publicly available at https://github.com/LeeJunHyun/NATR.

Utilizing market forecasts is pivotal in optimizing portfolio selection strategies. We introduce DeepClair, a novel framework for portfolio selection. DeepClair leverages a transformer-based time-series forecasting model to predict market trends, facilitating more informed and adaptable portfolio decisions. To integrate the forecasting model into a deep reinforcement learning-driven portfolio selection framework, we introduced a two-step strategy: first, pre-training the time-series model on market data, followed by fine-tuning the portfolio selection architecture using this model. Additionally, we investigated the optimization technique, Low-Rank Adaptation (LoRA), to enhance the pre-trained forecasting model for fine-tuning in investment scenarios. This work bridges market forecasting and portfolio selection, facilitating the advancement of investment strategies.

We propose a computational approach for recipe ideation, a downstream task that helps users select and gather ingredients for creating dishes. To perform this task, we developed RecipeMind, a food affinity score prediction model that quantifies the suitability of adding an ingredient to set of other ingredients. We constructed a large-scale dataset containing ingredient co-occurrence based scores to train and evaluate RecipeMind on food affinity score prediction. Deployed in recipe ideation, RecipeMind helps the user expand an initial set of ingredients by suggesting additional ingredients. Experiments and qualitative analysis show RecipeMind's potential in fulfilling its assistive role in cuisine domain.

Abstract Scientific novelty drives the efforts to invent new vaccines and solutions during the pandemic. First‐time collaboration and international collaboration are two pivotal channels to expand teams' search activities for a broader scope of resources required to address the global challenge, which might facilitate the generation of novel ideas. Our analysis of 98,981 coronavirus papers suggests that scientific novelty measured by the BioBERT model that is pretrained on 29 million PubMed articles, and first‐time collaboration increased after the outbreak of COVID‐19, and international collaboration witnessed a sudden decrease. During COVID‐19, papers with more first‐time collaboration were found to be more novel and international collaboration did not hamper novelty as it had done in the normal periods. The findings suggest the necessity of reaching out for distant resources and the importance of maintaining a collaborative scientific community beyond nationalism during a pandemic.

Several geographical latent representation models that capture geographical influences among points-of-interest (POIs) have been proposed. Although the models improve POI recommendation performance, they depend on shallow methods that cannot effectively capture highly non-linear geographical influences from complex user-POI networks. In this paper, we propose a new graph-based geographical latent representation model (GGLR) which can capture highly non-linear geographical influences from complex user-POI networks. Our proposed GGLR considers two types of geographical influences: ingoing influences and outgoing influences. Based on a graph auto-encoder, geographical latent representations of ingoing and outgoing influences are trained to increase geographical influences between two consecutive POIs that frequently appear in check-in histories. Furthermore, we propose a graph neural network-based POI recommendation model (GPR) that uses the trained geographical latent representations of ingoing and outgoing influences for the estimation of user preferences. In the experimental evaluation on real-world datasets, we show that GGLR effectively captures highly non-linear geographical influences and GPR achieves state-of-the-art performance.

Successive point-of-interest (POI) recommendation based on user check-in histories plays an important role in mobile-based social media platforms. Although a large amount of check-in data including textual content is generated from such platforms, most successive POI recommendation models do not leverage textual contents that provide useful information for understanding user interests. To address this problem, we propose a new content-aware successive POI recommendation (CAPRE) model in this paper. Based on a multi-head attention mechanism and a character-level convolutional neural network, CAPRE encodes usergenerated textual contents into content embedding to capture user interests. Based on long short-term memories (LSTMs), CAPRE capture content-aware user behavior patterns from encoded content embedding. Evaluation results on real-world datasets show that CAPRE achieves state-of-the-art recommendation performance.

Most companies utilize demographic information to develop their strategy in a market. However, such information is not available to most retail companies. Several studies have been conducted to predict the demographic attributes of users from their transaction histories, but they have some limitations. First, they focused on parameter sharing to predict all attributes but capturing task-specific features is also important in multi-task learning. Second, they assumed that all transactions are equally important in predicting demographic attributes. However, some transactions are more useful than others for predicting a certain attribute. Furthermore, decision making process of models cannot be interpreted as they work in a black-box manner. To address the limitations, we propose an Embedding Transformation Network with Attention (ETNA) model which shares representations at the bottom of the model structure and transforms them to task-specific representations using a simple linear transformation method. In addition, we can obtain more informative transactions for predicting certain attributes using the attention mechanism. The experimental results show that our model outperforms the previous models on all tasks. In our qualitative analysis, we show the visualization of attention weights, which provides business managers with some useful insights.

With the growing importance of personalized recommendation, numerous recommendation models have been proposed recently. Among them, Matrix Factorization (MF) based models are the most widely used in the recommendation field due to their high performance. However, MF based models suffer from cold start problems where user-item interactions are sparse. To deal with this problem, content based recommendation models which use the auxiliary attributes of users and items have been proposed. Since these models use auxiliary attributes, they are effective in cold start settings. However, most of the proposed models are either unable to capture complex feature interactions or not properly designed to combine user-item feedback information with content information. In this paper, we propose Self-Attentive Integration Network (SAIN) which is a model that effectively combines user-item feedback information and auxiliary information for recommendation task. In SAIN, a self-attention mechanism is used in the feature-level interaction layer to effectively consider interactions between multiple features, while the information integration layer adaptively combines content and feedback information. The experimental results on two public datasets show that our model outperforms the state-of-the-art models by 2.13%

With online calendar services gaining popularity worldwide, calendar data has become one of the richest context sources for understanding human behavior. However, event scheduling is still time-consuming even with the development of online calendars. Although machine learning based event scheduling models have automated scheduling processes to some extent, they often fail to understand subtle user preferences and complex calendar contexts with event titles written in natural language. In this paper, we propose Neural Event Scheduling Assistant (NESA) which learns user preferences and understands calendar contexts, directly from raw online calendars for fully automated and highly effective event scheduling. We leverage over 593K calendar events for NESA to learn scheduling personal events, and we further utilize NESA for multi-attendee event scheduling. NESA successfully incorporates deep neural networks such as Bidirectional Long Short-Term Memory, Convolutional Neural Network, and Highway Network for learning the preferences of each user and understanding calendar context based on natural languages. The experimental results show that NESA significantly outperforms previous baseline models in terms of various evaluation metrics on both personal and multi-attendee event scheduling tasks. Our qualitative analysis demonstrates the effectiveness of each layer in NESA and learned user preferences.

Opinion spam, intentionally written by spammers who do not have actual experience with services or products, has recently become a factor that undermines the credibility of information online. In recent years, studies have attempted to detect opinion spam using machine learning algorithms. However, limitations of gold-standard spam datasets still prove to be a major obstacle in opinion spam research. In this paper, we introduce a novel dataset called Paraphrased OPinion Spam (POPS), which contains a new type of review spam that imitates real human opinions using crowdsourcing. To create such a seemingly truthful review spam dataset, we asked task participants to paraphrase truthful reviews, and include factual information and domain knowledge in their reviews. The classification experiments and semantic analysis results show that our POPS dataset most linguistically and semantically resembles truthful reviews. We believe that our new deceptive opinion spam dataset will help advance opinion spam research.

User-generated content is becoming increasingly valuable to both individuals and businesses due to its usefulness and influence in e-commerce markets. As consumers rely more on such information, posting deceptive opinions, which can be deliberately used for potential profit, is becoming more of an issue. Existing work on opinion spam detection focuses mainly on linguistic features such as n-grams, syntactic patterns, or LIWC. However, deep semantic analysis remains largely unstudied. In this paper, we propose a frame-based deep semantic analysis method for understanding rich characteristics of deceptive and truthful opinions written by various types of individuals including crowdsourcing workers, employees who have expert-level domain knowledge about local businesses, and online users who post on Yelp and TripAdvisor. Using our proposed semantic frame feature, we developed a classification model that outperforms the baseline model and achieves an accuracy of nearly 91%. Also, we performed qualitative analysis of deceptive and truthful review datasets and considered their semantic differences. Finally, we successfully found some interesting features that existing methods were unable to identify.

Search engines have become an important decision making tool today. Decision making queries are often subjective, such as "a good birthday present for my girlfriend", "best action movies in 2010", to name a few. Unfortunately, such queries may not be answered properly by conventional search systems. In order to address this problem, we introduce Consento, a consensus search engine designed to answer subjective queries. Consento performs segment indexing, as opposed to document indexing, to capture semantics from user opinions more precisely. In particular, we define a new indexing unit, Maximal Coherent Semantic Unit (MCSU).

Schema matching is one of the key challenges in information integration. It is a labor-intensive and time-consuming process. To alleviate the problem, many automated solutions have been proposed. Most of the existing solutions mainly rely upon textual similarity of the data to be matched. However, there exist instances of the schema matching problem for which they do not even apply. Such problem instances typically arise when the column names in the schemas and the data in the columns are opaque or very difficult to interpret. In our previous work [36] we proposed a two-step technique to address this problem. In the first step, we measure the dependencies between attributes within tables using an information-theoretic measure and construct a dependency graph for each table capturing the dependencies among attributes. In the second step, we find matching node pairs across the dependency graphs by running a graph matching algorithm. In our previous work, we experimentally validated the accuracy of the approach. One remaining challenge is the computational complexity of the graph matching problem in the second step. In this paper we extend the previous work by improving the second phase of the algorithm incorporating efficient approximation algorithms into the framework.

We address the problem of handling topic oriented tasks on the World Wide Web. Our aim is to find most relevant and important pages for broad-topic queries while searching in a small set of candidate pages. We present a link analysis based algorithm SelHITS which is an improvement over Kleinberg's HITS algorithm. We introduce concept of virtual links to exploit latent information in the hyperlinked environment. Selective expansion of the root set and novel ranking strategy are the distinguishing features of our approach. Selective expansion method avoids topic drift and provides results consistent with only one interpretation of the query. Experimental evaluation and user feedback show that our algorithm indeed distills the most relevant and important pages for broad-topic queries. Trends in user feedback suggests that there exists a uniform notion of quality of search results within users

The entity resolution (ER) problem, which identifies duplicate entities that refer to the same real world entity, is essential in many applications. In this paper, in particular, we focus on resolving entities that contain a group of related elements in them (e.g., an author entity with a list of citations, a singer entity with song list, or an intermediate result by GROUP BY SQL query). Such entities, named as grouped-entities, frequently occur in many applications. The previous approaches toward grouped-entity resolution often rely on textual similarity, and produce a large number of false positives. As a complementing technique, in this paper, we present our experience of applying a recently proposed graph mining technique, Quasi-Clique, atop conventional ER solutions. Our approach exploits contextual information mined from the group of elements per entity in addition to syntactic similarity. Extensive experiments verify that our proposal improves precision and recall up to 83% when used together with a variety of existing ER solutions, but never worsens them.

We address the problem of answering broad-topic queries on the World Wide Web. We present a link based analysis algorithm SelHITS, which is an improvement over Kleinberg's HITS [2] algorithm. We introduce the concept of virtual links to exploit the latent information in the hyperlinked environment. We propose a novel approach to calculate hub and authority values. We also present a selective expansion method which avoids topic drift and provides results consistent with only one interpretation of the query, even if the query is ambiguous. Initial experimental evaluation and user feedback show that our algorithm indeed distills the most important and relevant pages for broad-topic queries. We also infer that there exists a uniform notion of quality of search results within users.

In this paper, we present a "value mapping" algorithm that does not rely on syntactic similarity or semantic interpretation of the values. The algorithm first constructs a statistical model (e.g., co-occurrence frequency or entropy vector) that captures the unique characteristics of values and their co-occurrence. It then finds the matching values by computing the distances between the models while refining the models using user feedback through iterations. Our experimental results suggest that our approach successfully establishes value mappings even in the presence of opaque data values and thus can be a useful addition to the existing data integration techniques.

We investigate algorithms for evaluating sliding window joins over pairs of unbounded streams. We introduce a unit-time-basis cost model to analyze the expected performance of these algorithms. Using this cost model, we propose strategies for maximizing the efficiency of processing joins in three scenarios. First, we consider the case where one stream is much faster than the other. We show that asymmetric combinations of join algorithms, (e.g., hash join on one input, nested-loops join on the other) can outperform symmetric join algorithm implementations. Second, we investigate the case where system resources are insufficient to keep up with the input streams. We show that we can maximize the number of join result tuples produced in this case by properly allocating computing resources across the two input streams. Finally, we investigate strategies for maximizing the number of result tuples produced when memory is limited, and show that proper memory allocation across the two input streams can result in significantly lower resource usage and/or more result tuples produced.

Most previous solutions to the schema matching problem rely in some fashion upon identifying "similar" column names in the schemas to be matched, or by recognizing common domains in the data stored in the schemas. While each of these approaches is valuable in many cases, they are not infallible, and there exist instances of the schema matching problem for which they do not even apply. Such problem instances typically arise when the column names in the schemas and the data in the columns are "opaque" or very difficult to interpret. In this paper we propose a two-step technique that works even in the presence of opaque column names and data values. In the first step, we measure the pair-wise attribute correlations in the tables to be matched and construct a dependency graph using mutual information as a measure of the dependency between attributes. In the second stage, we find matching node pairs in the dependency graphs by running a graph matching algorithm. We validate our approach with an experimental study, the results of which suggest that such an approach can be a useful addition to a set of (semi) automatic schema matching techniques.

The Strudel system applies concepts from database management systems to the process of building Web sites. Strudel's key idea is separating the management of the site's data, the creation and management of the site's structure, and the visual presentation of the site's pages. First, the site builder creates a uniform model of all data available at the site. Second, the builder uses this model to declaratively define the Web site's structure by applying a “site-definition query” to the underlying data. The result of evaluating this query is a “site graph”, which represents both the site's content and structure. Third, the builder specifies the visual presentation of pages in Strudel's HTML-template language. The data model underlying Strudel is a semi-structured model of labeled directed graphs.

article STRUDEL: a Web site management system Share on Authors: Mary Fernandez AT&T Labs, 600 Mountain Ave., Murray Hill, NJ AT&T Labs, 600 Mountain Ave., Murray Hill, NJView Profile , Daniela Florescu AT&T Labs, 600 Mountain Ave., Murray Hill, NJ AT&T Labs, 600 Mountain Ave., Murray Hill, NJView Profile , Jaewoo Kang AT&T Labs, 600 Mountain Ave., Murray Hill, NJ AT&T Labs, 600 Mountain Ave., Murray Hill, NJView Profile , Alon Levy AT&T Labs, 600 Mountain Ave., Murray Hill, NJ AT&T Labs, 600 Mountain Ave., Murray Hill, NJView Profile , Dan Suciu AT&T Labs, 600 Mountain Ave., Murray Hill, NJ AT&T Labs, 600 Mountain Ave., Murray Hill, NJView Profile Authors Info & Claims ACM SIGMOD RecordVolume 26Issue 2June 1997 pp 549–552https://doi.org/10.1145/253262.253403Online:01 June 1997Publication History 84citation759DownloadsMetricsTotal Citations84Total Downloads759Last 12 Months11Last 6 weeks1 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