Modeling Diverse Relevance Patterns in Ad-hoc Retrieval

By Yixing Fan, Jiafeng Guo, Yanyan Lan, Jun Xu, Chengxiang Zhai, Xueqi Cheng. 2018

94 citations recorded by the source corpus.

Assessing relevance between a query and a document is challenging in ad-hoc retrieval due to its diverse patterns, i.e., a document could be relevant to a query as a whole or partially as long as it provides sufficient information for users' need. Such diverse relevance patterns require an ideal retrieval model to be able to assess relevance in the right granularity adaptively. Unfortunately, most existing retrieval models compute relevance at a single granularity, either document-wide or passage-level, or use fixed combination strategy, restricting their ability in capturing diverse relevance patterns. In this work, we propose a data-driven method to allow relevance signals at different granularities to compete with each other for final relevance assessment. Specifically, we propose a HIerarchical Neural maTching model (HiNT) which consists of two stacked components, namely local matching layer and global decision layer. The local matching layer focuses on producing a set of local relevance signals by modeling the semantic matching between a query and each passage of a document. The global decision layer accumulates local signals into different granularities and allows them to compete with each other to decide the final relevance score.Experimental results demonstrate that our HiNT model outperforms existing state-of-the-art retrieval models significantly on benchmark ad-hoc retrieval datasets.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Interpretable & Time-Budget-Constrained Contextualization for Re-Ranking (2020, methodology): “Various approaches exist to reduce the match-matrix of term similarities to the matching score: using stacked Convolutional Neural Networks (CNN) [26, 22], parallel single-layered CNNs for n-gram interaction modelling [15], recurrent neural networks [8], and position independent counting methods.”
PROP: Pre-training with Representative Words Prediction for Ad-hoc Retrieval (2020, methodology): “Besides the above baselines, we also compare PROPwith existing state-of-the-art models (SOTA) on these five datasets, including CEDR-KNRM [21] on Robust04, BERT-maxP [10] on ClueWeb09-B, NWT [15] on Gov2, and HiNT [13] on MQ2007 and MQ2008.”
Leveraging Passage-level Cumulative Gain for Document Ranking (2020, methodology): “Other methods turned to deep neural networks to automatically learn the relationship [7, 29].”
Capreolus: A Toolkit for End-to-End Neural Ad Hoc Retrieval (2020, methodology): “interaction-focused models (DRMM [4], KNRM [16], PACRR [7], ConvKNRM [2], POSITDRMM [12]), hybrid models (DUET [11]), and hierarchical models (HiNT [3] and DeepTileBars [15]).”
A Deep Look into Neural Ranking Models for Information Retrieval (2019, methodology): “In academia, several neural ranking models learned from scratch can already outperform state-of-the-art LTR models with tens of hand-crafted features [33, 34].”
A Modern Perspective on Query Likelihood with Deep Generative Retrieval Models (2021, unknown): “Arguably, all existing neural ranking models [7, 11, 12, 14, 25, 28, 33, 50] follow the text matching paradigm, where relevance is calculated as the predicted matching score of each document to a given query.”
NPRF: A Neural Pseudo Relevance Feedback Framework for Ad-hoc Information Retrieval (2018, methodology): “Akin to the established neural IR models like DRMM (Guo et al., 2016), the description is based on a query-document pair, and a ranking can be produced by sorting the documents according to their scores.”
Local Self-Attention over Long Text for Efficient Document Retrieval (2020, unknown): “Several others [8, 14] have highlighted important considerations for designing neural ranking models for documents that are distinct from dealing with passages and other short text.”