The task of video grounding, which temporally localizes a natural language description in video, plays an important role understanding videos. Existing studies have adopted strategies sliding window over the entire or exhaustively ranking all possible clip-sentence pairs presegmented inevitably suffer from enumerated candidates. To alleviate this problem, we formulate as problem sequential decision making by learning agent regulates temporal grounding boundaries progressively based on its...

Image Retrieval is a fundamental task of obtaining images similar to the query one from database. A common image retrieval practice firstly retrieve candidate via similarity search using global features and then re-rank candidates by leveraging their local features. Previous learning-based studies mainly focus on either or representation learning tackle task. In this paper, we abandon two-stage paradigm seek design an effective single-stage solution integrating information inside into...

The task of travel time estimation (TTE), which estimates the for a given route and departure time, plays an important role in intelligent transportation systems such as navigation, planning, ride-hailing services. This is challenging because many essential aspects, traffic prediction contextual information. First, accuracy strongly correlated with speed road segments route. Existing work mainly adopts spatial-temporal graph neural networks to improve prediction, where spatial temporal...

The constrained outbreak of COVID-19 in Mainland China has recently been regarded as a successful example fighting this highly contagious virus. Both the short period (in about three months) transmission and sub-exponential increase confirmed cases have proved that Chinese authorities took effective epidemic prevention measures, such case isolation, travel restrictions, closing recreational venues, banning public gatherings. These measures can, course, effectively control spread pandemic....

Next point-of-interest (POI) recommendation is a hot research field where recent emerging scenario, next POI to search recommendation, has been deployed in many online map services such as Baidu Maps. One of the key issues this scenario providing satisfactory for cold-start cities with limited number user-POI interactions, which requires transferring knowledge hidden rich data from other these cities. Existing literature either does not consider city-transfer issue or cannot simultaneously...

Abstract The rising mental health difficulties of the urban population in developing countries may be attributed to high levels air pollution. However, nationwide large-scale empirical works that examine this claim are rare. In study, we construct a daily metric using volume mental-health-related queries on largest search engine China, Baidu, test hypothesis. We find pollution causally undermines people’s and impact becomes stronger as duration exposure increases. Heterogeneity analyses...

Pre-trained models (PTMs) have become a fundamental backbone for downstream tasks in natural language processing and computer vision. Despite initial gains that were obtained by applying generic PTMs to geo-related at Baidu Maps, clear performance plateau over time was observed. One of the main reasons this is lack readily available geographic knowledge PTMs. To address problem, paper, we present ERNIE-GeoL, which geography-and-language pre-trained model designed developed improving Maps....

The increasing interest in international travel has raised the demand of retrieving point interests (POIs) multiple languages. This is even superior to find local venues such as restaurants and scenic spots unfamiliar languages when traveling abroad. Multilingual POI retrieval, enabling users desired POIs a demanded language using queries numerous languages, become an indispensable feature today's global map applications Baidu Maps. task non-trivial because two key challenges: (1) visiting...

Travel time estimation (TTE) is one of the most critical modules at Baidu Maps, which plays a vital role in intelligent transportation services such as route planning and navigation. During driving en route, navigation system Maps can provide real-time estimations on when user will arrive destination. It automatically recalculates updates remaining travel from driver's current position to destination (hereafter referred route) every few minutes. The previously deployed TTE model i.e.,...

The task of road extraction has aroused remarkable attention due to its critical role in facilitating urban development and up-to-date map maintenance, which widespread applications such as navigation autonomous driving. Existing solutions either rely on a single source data for graph or simply fuse the multimodal information sub-optimal way. In this paper, we present an automatic solution named DuARE, is designed exploit knowledge underlying fully manner. Specifically, collect large-scale...

Urban villages (UVs) refer to the underdeveloped informal settlement falling behind rapid urbanization in a city. Since there are high levels of social inequality and risks these UVs, it is critical for city managers discover all UVs making appropriate renovation policies. Existing approaches detecting labor-intensive or have not fully addressed unique challenges UV detection such as scarcity labeled diverse urban patterns different regions. To this end, we first build an region graph (URG)...

Entity recommendation, providing search users with an improved experience by assisting them in finding related entities for a given query, has become indispensable feature of today's Web engine. Existing studies typically only consider the query issued at current time step while ignoring in-session preceding queries. Thus, they fail to handle ambiguous queries such as "apple" because model could not understand which apple (company or fruit) is talked about. In this work, we believe that...

While recommender systems have been ubiquitously used in digital marketing and online business development, the conversions of advertising for mobile apps installation activation sometimes are far from satisfactory, due to lack feedback App-related activities, leading a poor record Return on Investment (RoI). Though advertisers, e.g., App operators Store, granted log users' app-related activities such as installation, activation, usages, preferences per agreement, they usually limit access...

Understanding urban regional characteristics is pivotal in driving critical insights for planning and management. We have witnessed the successful application of pre-trained Foundation Models (FMs) generating universal representations various downstream tasks. However, applying this principle to geospatial domain remains challenging, primarily due difficulty gathering extensive data developing a dedicated foundation model. Though there been some attempts empower existing FMs with data, most...

An up-to-date city-scale lane-level map is an indispensable infrastructure and a key enabling technology for ensuring the safety user experience of autonomous driving systems.In industrial scenarios, reliance on manual annotation updates creates critical bottleneck.Lane-level require precise change information must ensure consistency with adjacent data while adhering to strict standards.Traditional methods utilize three-stage approach-construction, detection, updating-which often...

We study ``selective'' or ``conditional'' classification problems under an agnostic setting. Classification tasks commonly focus on modeling the relationship between features and categories that captures vast majority of data. In contrast to common machine learning frameworks, conditional intends model such relationships only a subset data defined by some selection rule. Most work either solves problem in realizable setting does not guarantee error is bounded compared optimal solution. this...

Entity recommendation, providing entity suggestions to assist users in discovering interesting information, has become an indispensable feature of today’s Web search engine. However, the majority existing recommendation methods are not designed boost performance terms serendipity, which also plays important role appreciation for a system. To keep engaged, it is take into account serendipity when building In this article, we propose learning recommend framework that consists two components:...

Point of interest auto-completion (POI-AC) is a featured function in the search engine many Web mapping services. This keeps suggesting dynamic list POIs as user types each character, and it can dramatically save effort typing, which quite useful on mobile devices. Existing approaches POI-AC for industrial use mainly adopt various learning to rank (LTR) models with handcrafted features even historically clicked are taken into account personalization. However, these prior arts tend reach...

To contain the pandemic of coronavirus (COVID-19) in Mainland China, authorities have put place a series measures, including quarantines, social distancing, and travel restrictions. While these strategies effectively dealt with critical situations outbreaks, combination mobility controls has slowed China's economic growth, resulting first quarterly decline Gross Domestic Product (GDP) since GDP began to be calculated, 1992. characterize potential shrinkage domestic economy, from perspective...

The novel coronavirus disease (COVID-19) has crushed daily routines and is still rampaging through the world. Existing solution for nonpharmaceutical interventions usually needs to timely precisely select a subset of residential urban areas containment or even quarantine, where spatial distribution confirmed cases been considered as key criterion selection. While such measure successfully stopped slowed down spread COVID-19 in some countries, it criticized being inefficient ineffective,...

Point Of Interest Auto-Completion (abbr. as POI-AC) is one of the featured functions for search engine at Baidu Maps. It can dynamically suggest a list POI candidates within milliseconds user enters each character (e.g., English, Chinese, or Pinyin character) into box. Ideally, may need to provide only and immediately obtain desired top suggested by POI-AC. In this way, user's keystrokes be dramatically saved, which significantly reduces time effort typing, especially on mobile devices that...

Estimated time of arrival (ETA) prediction, also known as travel estimation, is a fundamental task for wide range intelligent transportation applications, such navigation, route planning, and ride-hailing services. To accurately predict the route, it essential to take into account both contextual predictive factors, spatial-temporal interaction, driving behavior, traffic congestion propagation inference. The ETA prediction models previously deployed at Baidu Maps have addressed factors...

Providing a plausible explanation for the relationship between two related entities is an important task in some applications of knowledge graphs, such as search engines. However, most existing methods require large number manually labeled training data, which cannot be applied large-scale graphs due to expensive data annotation. In addition, these typically rely on costly handcrafted features. this paper, we propose effective pairwise ranking model by leveraging clickthrough Web engine...