The cold-start problem remains a significant challenge in recommendation systems, particularly for new users or unseen items with little to no historical data. Existing methods, including graph neural networks, often struggle in such scenarios. Inspired by the success of transformer models in natural language processing, we propose G-TRAC (Graph-Textual Representations Alignment for Cold-start Recommendations), a novel approach that integrates transformer-based textual modeling with graph neural networks. By effectively leveraging both textual and structural information, G-TRAC addresses cold-start challenges more effectively. Extensive experiments demonstrate its ability to enhance recommendation quality and generalize well across diverse scenarios.

Modern information retrieval (IR) systems, powered by bi-encoder architectures and pretrained language models, rely on effective negative sampling for contrastive learning. While easy negatives are computationally simple, they fail to challenge the model, whereas hard negatives-selected via methods like BM25, ANCE, or ADORE-can be overly difficult and misleading. To this end, this paper proposes dynamic margin-based contrastive learning (DMCL), which adaptively adjusts the decision boundary based on query-negative similarity, ensuring consistent exposure to moderately hard negatives. Experiments across diverse datasets and models show that DMCL outperforms traditional methods, achieving state-of-the-art retrieval performance with minimal computational cost.

This paper introduces SARA, a semantic-assisted reinforced active learning framework for enhancing entity alignment (EA) under limited supervision scenarios. SARA addresses the challenges of EA in real-world scenarios, including knowledge graph heterogeneity and limited training ground truth. SARA effectively selects valuable entity pairs with limited labeled data by combining reinforced active learning and semantic information. It utilizes a pair-wise language model based on Sentence-BERT to learn informative name embeddings that capture entity name semantics. These embeddings are combined with structural embeddings and trained using a novel semantic-assisted alignment loss. Extensive experiments on benchmark datasets and a real-world dataset demonstrate the superiority of SARA over existing approaches, particularly in limited labeled data scenarios. The paper also provides insights into fine-tuning strategies, presents ablation studies, and conducts sensitivity analyses to validate the effectiveness of SARA.

This paper presents ConvRerank, a conversational passage re-ranker that employs a newly developed pseudo-labeling approach. Our proposed view-ensemble method enhances the quality of pseudo-labeled data, thus improving the fine-tuning of ConvRerank. Our experimental evaluation on benchmark datasets shows that combining ConvRerank with a conversational dense retriever in a cascaded manner achieves a good balance between effectiveness and efficiency. Compared to baseline methods, our cascaded pipeline demonstrates lower latency and higher top-ranking effectiveness. Furthermore, the in-depth analysis confirms the potential of our approach to improving the effectiveness of conversational search.

Prediction using the ground truth sounds like an oxymoron in machine learning. However, such an unrealistic setting was used in hundreds, if not thousands of papers in the area of finding graph representations. To evaluate the multi-label problem of node classification by using the obtained representations, many works assume that the number of labels of each test instance is known in the prediction stage. In practice such ground truth information is rarely available, but we point out that such an inappropriate setting is now ubiquitous in this research area. We detailedly investigate why the situation occurs. Our analysis indicates that with unrealistic information, the performance is likely over-estimated. To see why suitable predictions were not used, we identify difficulties in applying some multi-label techniques. For the use in future studies, we propose simple and effective settings without using practically unknown information. Finally, we take this chance to compare major graph representation learning methods on multi-label node classification.

In this demonstration, we present RecDelta, an interactive tool for the cross-model evaluation of top-k recommendation. RecDelta is a web-based information system where people visually compare the performance of various recommendation algorithms and their recommended items. In the proposed system, we visualize the distribution of the δ scores between algorithms--a distance metric measuring the intersection between recommendation lists. Such visualization allows for rapid identification of users for whom the items recommended by different algorithms diverge or vice versa; then, one can further select the desired user to present the relationship between recommended items and his/her historical behavior. RecDelta benefits both academics and practitioners by enhancing model explainability as they develop recommendation algorithms with their newly gained insights. Note that while the system is now online at https://cfda.csie.org/recdelta, we also provide a video recording at https://tinyurl.com/RecDelta to introduce the concept and the usage of our system.

Explicit feedback---user input regarding their interest in an item---is the most helpful information for recommendation as it comes directly from the user and shows their direct interest in the item. Most approaches either treat the recommendation given such feedback as a typical regression problem or regard such data as implicit and then directly adopt approaches for implicit feedback; both methods, however,tend to yield unsatisfactory performance in top-k recommendation. In this paper, we propose interaction-level preference ranking(IPR), a novel pairwise ranking embedding learning approach to better utilize explicit feedback for recommendation. Experiments conducted on three real-world datasets show that IPR yields the best results compared to six strong baselines.

Item concept modeling is commonly achieved by leveraging textual information. However, many existing models do not leverage the inferential property of concepts to capture word meanings, which therefore ignores the relatedness between correlated concepts, a phenomenon which we term conceptual “correlation sparsity.” In this paper, we distinguish between word modeling and concept modeling and propose an item concept modeling framework centering around the item concept network (ICN). ICN models and further enriches item concepts by leveraging the inferential property of concepts and thus addresses the correlation sparsity issue. Specifically, there are two stages in the proposed framework: ICN construction and embedding learning. In the first stage, we propose a generalized network construction method to build ICN, a structured network which infers expanded concepts for items via matrix operations. The second stage leverages neighborhood proximity to learn item and concept embeddings. With the proposed ICN, the resulting embedding facilitates both homogeneous and heterogeneous tasks, such as item-to-item and concept-to-item retrieval, and delivers related results which are more diverse than traditional keyword-matching-based approaches. As our experiments on two real-world datasets show, the framework encodes useful conceptual information and thus outperforms traditional methods in various item classification and retrieval tasks.

In this demonstration, we develop an interactive tool, HIVE, to demonstrate the ability and versatility of an explainable risk ranking model with a special focus on financial use cases. HIVE is a web-based tool that provides users with automated highlighted financial statements, and HIVE is designed for making comparing statements rather more efficient. Moreover, with the proposed tool, users can find related reports at ease, and we believe that HIVE can benefit both academics and practitioners in finance as they can work around deep learning models with their newly gained insights.

Considering the temporal order of user-item interactions for recommendation forms a novel class of recommendation algorithms in recent years, among which sequential recommendation models are the most popular approaches. Although, theoretically, such fine-grained modeling should be beneficial to the recommendation performance, these sequential models in practice greatly suffer from the issue of data sparsity as there are a huge number of combinations for item sequences. To address the issue, we propose LSTPR, a graph-based matrix factorization model that incorporates both high-order graph information and long short-term user preferences into the modeling process. LSTPR explicitly distinguishes long-term and short-term user preferences and enriches the sparse interactions via random surfing on the user-item graph. Experiments on three recommendation datasets with temporal user-item information demonstrate that the proposed LSTPR model achieves significantly better performance than the seven baseline methods.

Mood and emotion play an important role when it comes to choosing musical tracks to listen to. In the field of music information retrieval and recommendation, emotion is considered contextual information that is hard to capture, albeit highly influential. In this study, we analyze the connection between users` emotional states and their musical choices. Particularly, we perform a large-scale study based on two data sets containing 560,000 and 90,000 #nowplaying tweets, respectively. We extract affective contextual information from hashtags contained in these tweets by applying an unsupervised sentiment dictionary approach. Subsequently, we utilize a state-of-the-art network embedding method to learn latent feature representations of users, tracks and hashtags. Based on both the affective information and the latent features, a set of eight ranking methods is proposed. We find that relying on a ranking approach that incorporates the latent representations of users and tracks allows for capturing a user's general musical preferences well (regardless of used hashtags or affective information). However, for capturing context-specific preferences (a more complex and personal ranking task), we find that ranking strategies that rely on affective information and that leverage hashtags as context information outperform the other ranking strategies.

Conversational search plays a vital role in conversational information seeking. As queries in information seeking dialogues are ambiguous for traditional ad hoc information retrieval (IR) systems due to the coreference and omission resolution problems inherent in natural language dialogue, resolving these ambiguities is crucial. In this article, we tackle conversational passage retrieval, an important component of conversational search, by addressing query ambiguities with query reformulation integrated into a multi-stage ad hoc IR system. Specifically, we propose two conversational query reformulation (CQR) methods: (1) term importance estimation and (2) neural query rewriting. For the former, we expand conversational queries using important terms extracted from the conversational context with frequency-based signals. For the latter, we reformulate conversational queries into natural, stand-alone, human-understandable queries with a pretrained sequence-to-sequence model. Detailed analyses of the two CQR methods are provided quantitatively and qualitatively, explaining their advantages, disadvantages, and distinct behaviors. Moreover, to leverage the strengths of both CQR methods, we propose combining their output with reciprocal rank fusion, yielding state-of-the-art retrieval effectiveness, 30% improvement in terms of NDCG@3 compared to the best submission of Text REtrieval Conference (TREC) Conversational Assistant Track (CAsT) 2019.

Textual data is common and informative auxiliary information for recommender systems. Most prior art utilizes text for rating prediction, but rare work connects it to top-recommendation. Moreover, although advanced recommendation models capable of incorporating auxiliary information have been developed, none of these are specifically designed to model textual information, yielding a limited usage scenario for typical user-to-item recommendation. In this work, we present a framework of text-aware preference ranking (TPR) for top- recommendation, in which we comprehensively model the joint association of user-item interaction and relations between items and associated text. Using the TPR framework, we construct a joint likelihood function that explicitly describes two ranking structures: 1) item preference ranking (IPR) and 2) word relatedness ranking (WRR), where the former captures the item preference of each user and the latter captures the word relatedness of each item. As these two explicit structures are by nature mutually dependent, we propose TPR-OPT, a simple yet effective learning criterion that additionally includes implicit structures, such as relatedness between items and relatedness between words for each user for model optimization. Such a design not only successfully describes the joint association among users, words, and text comprehensively but also naturally yields powerful representations that are suitable for a range of recommendation tasks, including user-to-item, item-to-item, and user-to-word recommendation, as well as item-to-word reconstruction. In this paper, extensive experiments have been conducted on eight recommendation datasets, the results of which demonstrate that by including textual information from item descriptions, the proposed TPR model consistently outperforms state-of-the-art baselines on various recommendation tasks.

While internalized "implicit knowledge" in pretrained transformers has led to fruitful progress in many natural language understanding tasks, how to most effectively elicit such knowledge remains an open question.Based on the text-to-text transfer transformer (T5) model, this work explores a template-based approach to extract implicit knowledge for commonsense reasoning on multiple-choice (MC) question answering tasks.Experiments on three representative MC datasets show the surprisingly good performance of our simple template, coupled with a logit normalization technique for disambiguation.Furthermore, we verify that our proposed template can be easily extended to other MC tasks with contexts such as supporting facts in open-book question answering settings.Starting from the MC task, this work initiates further research to find generic natural language templates that can effectively leverage stored knowledge in pretrained models.

We propose the use of a word-level sentiment bidirectional LSTM in tandem with the self-attention mechanism for sentence-level sentiment prediction. In addition to the proposed model, we also present a finance report dataset for sentence-level financial risk detection. Experiments conducted on the proposed dataset together with two public review datasets attest the effectiveness of our model for sentence sentiment prediction.

In this paper, we propose a novel optimization criterion that leverages features of the skew normal distribution to better model the problem of personalized recommendation. Specifically, the developed criterion borrows the concept and the flexibility of the skew normal distribution, based on which three hyperparameters are attached to the optimization criterion. Furthermore, from a theoretical point of view, we not only establish the relation between the maximization of the proposed criterion and the shape parameter in the skew normal distribution, but also provide the analogies and asymptotic analysis of the proposed criterion to maximization of the area under the ROC curve. Experimental results conducted on a range of large-scale real-world datasets show that our model significantly outperforms the state of the art and yields consistently best performance on all tested datasets.

We present FRIDAYS, a financial risk information detecting and analyzing system that enables financial professionals to efficiently comprehend financial reports in terms of risk and domain-specific sentiment cues. Our system is designed to integrate multiple NLP models trained on financial reports but on different levels (i.e., word, multi-word, and sentence levels) and to illustrate the prediction results generated by the models. The system is available online at https://cfda.csie.org/FRIDAYS/.

Customer reviews on platforms such as TripAdvisor and Amazon provide rich information about the ways that people convey sentiment on certain domains. Given these kinds of user reviews, this paper proposes UGSD, a representation learning framework for constructing domain-specific sentiment dictionaries from online customer reviews, in which we leverage the relationship between user-generated reviews and the ratings of the reviews to associate the reviewer sentiment with certain entities. The proposed framework has the following three main advantages. First, no additional annotations of words or external dictionaries are needed for the proposed framework; the only resources needed are the review texts and entity ratings. Second, the framework is applicable across a variety of user-generated content from different domains to construct domain-specific sentiment dictionaries. Finally, each word in the constructed dictionary is associated with a low-dimensional dense representation and a degree of relatedness to a certain rating, which enable us to obtain more fine-grained dictionaries and enhance the application scalability of the constructed dictionaries as the word representations can be adopted for various tasks or applications, such as entity ranking and dictionary expansion. The experimental results on three real-world datasets show that the framework is effective in constructing high-quality domain-specific sentiment dictionaries from customer reviews.

This paper attempts to conduct a sentence-level sentiment analysis with respect to financial risk on a collection of financial reports. Specifically, we first propose a simple yet efficient algorithm to generate financial sentiment phrases (senti-phrases), and then with the obtained senti-phrases, we utilize multiple sentence embedding models for better learning the representations of financial risk sentences. In order to verify the performance of the proposed approach, we conduct a risk classification task of financial sentences on a sentence-level labeled dataset of finance reports. Experimental results show that incorporating the obtained senti-phrases into the embedding-based models improves the classification performance.

In the Age of Big Data, graph embedding has received increasing attention for its ability to accommodate the explosion in data volume and diversity, which challenge the foundation of modern recommender systems. Respectively, graph facilitates fusing complex systems of interactions into a unified structure and distributed embedding enables efficient retrieval of entities, as in the case of approximate nearest neighbor (ANN) search. When combined, graph embedding captures relational information beyond entity interaction and towards a problem's underlying structure, as epitomized by struct2vec [20] and PinSage [26]. This session will start by brushing up on the basics about graphs and embedding methods and discussing their merits. We then quickly dive into using the mathematical formulation of graph embedding to derive the modular framework: Sampler-Mapper-Optimizer for Recommendation, or SMORe. We demonstrate existing models used for recommendation, such as MF and BPR, can all be assembled using three basic components: sampler, mapper, and optimizer. The tutorial is accompanied by a hands-on session, where we show how graph embedding can model complex systems through the multi-task learning and the cross-platform data sparsity alleviation tasks.

We present collaborative similarity embedding (CSE), a unified framework that exploits comprehensive collaborative relations available in a user-item bipartite graph for representation learning and recommendation. In the proposed framework, we differentiate two types of proximity relations: direct proximity and k-th order neighborhood proximity. While learning from the former exploits direct user-item associations observable from the graph, learning from the latter makes use of implicit associations such as user-user similarities and item-item similarities, which can provide valuable information especially when the graph is sparse. Moreover, for improving scalability and flexibility, we propose a sampling technique that is specifically designed to capture the two types of proximity relations. Extensive experiments on eight benchmark datasets show that CSE yields significantly better performance than state-of-the-art recommendation methods.

Recommender systems are vital ingredients for many e-commerce services. In the literature, two of the most popular approaches are based on factorization and graph-based models; the former approach captures user preferences by factorizing the observed direct interactions between users and items, and the latter extracts indirect preferences from the graphs constructed by user-item interactions. In this paper we present HOP-Rec, a unified and efficient method that incorporates the two approaches. The proposed method involves random surfing on a graph to harvest high-order information among neighborhood items for each user. Instead of factorizing a transition matrix, our method introduces a confidence weighting parameter to simulate all high-order information simultaneously, for which we maintain a sparse user-item interaction matrix and enrich the matrix for each user using random walks. Experimental results show that our approach significantly outperforms the state of the art on a range of large-scale real-world datasets.

We present NavWalker, a flexible random walk-based approach for learning the representations of vertices in an information network. The proposed method enables us to incorporate different walk strategies into the sampling process of random walks, in order to further boost the network embedding techniques. Specifically, we formulate the proposed method by integrating the adjacency matrix of a network with a pre-defined information augmentation matrix. In contrast to SkipGram-based network embedding methods such as DeepWalk and Node2vec, which use only local network information to learn the representations, our method is flexible to further incorporate global or other auxiliary network information to guide the sampling process. Experiments on six real-world datasets demonstrate the advantages of the flexibility and its superior performance as compared to other state-of-the-art network embedding algorithms for the tasks of classification and recommendation.

This paper attempts to conduct analysis for one certain type of user reviews; that is, the reviews on a super-entity (e.g., restaurant) involve descriptions for many sub-entities (e.g., dishes). To deal with such analysis, we propose a text embedding framework for ranking sub-entities from user reviews of a given super-entity. Experiments on two real-world datasets show that our method outperforms three baselines by a statistically significant amount. Intriguing cases from the experiments are discussed in the paper.

This paper proposes an item concept embedding (ICE) framework to model item concepts via textual information. Specifically, in the proposed framework there are two stages: graph construction and embedding learning. In the first stage, we propose a generalized network construction method to build a network involving heterogeneous nodes and a mixture of both homogeneous and heterogeneous relations. The second stage leverages the concept of neighborhood proximity to learn the embeddings of both items and words. With the proposed carefully designed ICE networks, the resulting embedding facilitates both homogeneous and heterogeneous retrieval, including item-to-item and word-to-item retrieval. Moreover, as a distributed embedding approach, the proposed ICE approach not only generates related retrieval results but also delivers more diverse results than traditional keyword-matching-based approaches. As our experiments on two real-world datasets show, ICE encodes useful textual information and thus outperforms traditional methods in various item classification and retrieval tasks.

In this demonstration, we present FIN10K, a web-based information system that facilitates the analysis of textual information in financial reports. The proposed system has three main components: (1) a 10-K Corpus, including an inverted index of financial reports on Form 10-K, several numerical finance measures, and pre-trained word embeddings; (2) an information retrieval system; and (3) two data visualizations of the analyzed results. The system can be of great help in revealing valuable insights within large amounts of textual information. The system is now online available at http: //clip.csie.org/10K/.

A common scenario considered in recommender systems is to predict a user's preferences on unseen items based on his/her preferences on observed items. A major limitation of this scenario is that a user might be interested in different things each time when using the system, but there is no way to allow the user to actively alter or adjust the recommended results. To address this issue, we propose the idea of "query-based recommendation" that allows a user to specify his/her search intention while exploring new items, thereby incorporating the concept of information retrieval into recommendation systems. Moreover, the idea is more desirable when the user intention can be expressed in different ways. Take music recommendation as an example: the proposed system allows a user to explore new song tracks by specifying either a track, an album, or an artist. To enable such heterogeneous queries in a recommender system, we present a novel technique called "Heterogeneous Preference Embedding" to encode user preference and query intention into low-dimensional vector spaces. Then, with simple search methods or similarity calculations, we can use the encoded representation of queries to generate recommendations. This method is fairly flexible and it is easy to add other types of information when available. Evaluations on three music listening datasets confirm the effectiveness of the proposed method over the state-of-the-art matrix factorization and network embedding methods.

This paper proposes to apply the contin-uous vector representations of words for discovering keywords from a financial sen-timent lexicon. In order to capture more keywords, we also incorporate syntactic information into the Continuous Bag-of-Words (CBOW) model. Experimental re-sults on a task of financial risk prediction using the discovered keywords demonstrate that the proposed approach is good at pre-dicting financial risk. 1

Due to more and more request of visual entertainment with improved perceptual realism, stereo video offering enhanced realism and interactivity is highly desired. Usually, the stereo video is packed into a single video and each view preserves only half resolution to achieve an efficient delivery. In this paper, a novel frame-compatible stereo video coding technique based on content-aware video retargeting is presented. Despite of combing the stereo video into a single channel by uniform downsampling, we propose to sub-sample the stereo video in an unequal way, by taking the saliency map of the video into considering. During the reconstruction of the stereo video with original resolution, the regions with higher saliency attention are less distorted, relying on a smaller downsampling factor, instead of 2. The experimental results reveal that the proposed technique achieves up to 43% bitrate saving as compared to the most popular frame-compatible stereo formats.

This paper proposes a context-aware approach that recommends music to a user based on the user's emotional state predicted from the article the user writes. We analyze the association between user-generated text and music by using a real-world dataset with user, text, music tripartite information collected from the social blogging website LiveJournal. The audio information represents various perceptual dimensions of music listening, including danceability, loudness, mode, and tempo; the emotional text information consists of bag-of-words and three dimensional affective states within an article: valence, arousal and dominance. To combine these factors for music recommendation, a factorization machine-based approach is taken. Our evaluation shows that the emotional context information mined from user-generated articles does improve the quality of recommendation, comparing to either the collaborative filtering approach or the content-based approach.

This paper proposes a music recommendation approach based on various similarity information via Factorization Machines (FM). We introduce the idea of similarity, which has been widely studied in the filed of information retrieval, and incorporate multiple feature similarities into the FM framework, including content-based and context-based similarities. The similarity information not only captures the similar patterns from the referred objects, but enhances the convergence speed and accuracy of FM. In addition, in order to avoid the noise within large similarity of features, we also adopt the grouping FM as an extended method to model the problem. In our experiments, a music-recommendation dataset is used to assess the performance of the proposed approach. The datasets is collected from an online blogging Web site, which includes user listening history, user profiles, social information, and music information. Our experimental results show that, with various types of feature similarities the performance of music recommendation can be enhanced significantly. Furthermore, via the grouping technique, the performance can be improved significantly in terms of Mean Average Precision, compared to the traditional collaborative filtering approach.

With the significant advancement of statistical timing and yield analysis algorithms, there is a strong need for accurate and analytical spatial correlation models. In this paper, we propose a novel spatial correlation modeling method that can not only capture the general spatial correlation relationship but also can generate highly accurate and analytical models. Our method, based on singular value decomposition, can generate sequences of polynomial weighted by the singular values. Experimental results from foundry measurement data show that our proposed approach is 3x accuracy improvement over several distance based spatial correlation modeling methods.

With the significant advancement of statistical timing and yield analysis algorithms, there is a strong need for accurate and analytical spatial correlation models. In this paper, we propose a novel spatial correlation modeling method not only can capture the general spatial correlation relationship but also can generate highly accurate and analytical models. Our method, based on Singular Value Decomposition (SVD), can generate sequences of polynomial weighted by the singular values. Experimental results from foundry measurement data show that our proposed approach is 5X accuracy improvement over several distance based spatial correlation modeling methods.

Most of the thyroid nodules are heterogeneous with various internal components, which confuse many radiologists and physicians with their various echo patterns in thyroid nodules. A lot of texture extraction methods were used to characterize the thyroid nodules. Accordingly, the thyroid nodules could be classified by the corresponding textural features. In this paper, five support vector machines (SVM) were adopted to select the significant textural features and to classify the nodular lesions of thyroid. Experimental results showed the proposed method classifies the thyroid nodules correctly and efficiently. The comparison results demonstrated that the capability of feature selection of the proposed method was similar to the sequential floating forward selection (SFFS) method. However, the proposed method is faster than the SFFS method.

This paper proposes a learning approach for the merging process in multilingual information retrieval (MLIR). To conduct the learning approach, we also present a large number of features that may influence the MLIR merging process; these features are mainly extracted from three levels: query, document, and translation. After the feature extraction, we then use the FRank ranking algorithm to construct a merge model; to our knowledge, this practice is the first attempt to use a learning-based ranking algorithm to construct a merge model for MLIR merging. In our experiments, three test collections for the task of crosslingual information retrieval (CLIR) in NTCIR3, 4, and 5 are employed to assess the performance of our proposed method; moreover, several merging methods are also carried out for a comparison, including traditional merging methods, the 2-step merging strategy, and the merging method based on logistic regression. The experimental results show that our method can significantly improve merging quality on two different types of datasets. In addition to the effectiveness, through the merge model generated by FRank, our method can further identify key factors that influence the merging process; this information might provide us more insight and understanding into MLIR merging.

The paper is concerned with learning to rank, which is to construct a model or a function for ranking objects. Learning to rank is useful for document retrieval, collaborative filtering, and many other applications. Several methods for learning to rank have been proposed, which take object pairs as 'instances' in learning. We refer to them as the pairwise approach in this paper. Although the pairwise approach offers advantages, it ignores the fact that ranking is a prediction task on list of objects. The paper postulates that learning to rank should adopt the listwise approach in which lists of objects are used as 'instances' in learning. The paper proposes a new probabilistic method for the approach. Specifically it introduces two probability models, respectively referred to as permutation probability and top k probability, to define a listwise loss function for learning. Neural Network and Gradient Descent are then employed as model and algorithm in the learning method. Experimental results on information retrieval show that the proposed listwise approach performs better than the pairwise approach.

Ranking problem is becoming important in many fields, especially in information retrieval (IR). Many machine learning techniques have been proposed for ranking problem, such as RankSVM, RankBoost, and RankNet. Among them, RankNet, which is based on a probabilistic ranking framework, is leading to promising results and has been applied to a commercial Web search engine. In this paper we conduct further study on the probabilistic ranking framework and provide a novel loss function named fidelity loss for measuring loss of ranking. The fidelity loss notonly inherits effective properties of the probabilistic ranking framework in RankNet, but possesses new properties that are helpful for ranking. This includes the fidelity loss obtaining zero for each document pair, and having a finite upper bound that is necessary for conducting query-level normalization. We also propose an algorithm named FRank based on a generalized additive model for the sake of minimizing the fedelity loss and learning an effective ranking function. We evaluated the proposed algorithm for two datasets: TREC dataset and real Web search dataset. The experimental results show that the proposed FRank algorithm outperforms other learning-based ranking methods on both conventional IR problem and Web search.