Assessment of learning to rank methods for query expansion

Xu, Bo; Lin, Hongfei; Lin, Yuan

doi:10.1002/asi.23476

Cited by 18 publications

(19 citation statements)

References 30 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The features used to make the selection include querydependent features, making the selection query dependent. More recently, Xu et al [77] proposed a learning-to-rank based query expansion. It re-ranks a set of potential terms for expanding query by exploiting the top retrieved documents and the collection statistics.…”

Section: Adaptive Systemsmentioning

confidence: 99%

Learning to Adaptively Rank Document Retrieval System Configurations

Deveaud

Mothe

Ullah

et al. 2018

ACM Trans. Inf. Syst.

View full text Add to dashboard Cite

Modern Information Retrieval (IR) systems have become more and more complex, involving a large number of parameters. For example, a system may choose from a set of possible retrieval models (BM25, language model, etc.), or various query expansion parameters, whose values greatly in uence the overall retrieval effectiveness. Traditionally, these parameters are set at a system level based on training queries, and the same parameters are then used for di erent queries. We observe that it may not be easy to set all these parameters separately, since they can be dependent. In addition, a global setting for all queries may not best t all individual queries with di erent characteristics. The parameters should be set according to these characteristics. In this article, we propose a novel approach to tackle this problem by dealing with the entire system con gurations (i.e., a set of parameters representing a n IR system b ehaviour) instead of selecting a single parameter at a time. The selection of the best con guration i s c ast a s a p roblem o f r anking di erent possible con gurations given a query. We apply learning-to-rank approaches for this task. We exploit both the query features and the system con guration f eatures i n t he l earning-to-rank m ethod s o t hat the selection of con guration i s q uery d ependent. T he e xperiments w e c onducted o n f our T REC a d h oc collections show that this approach can signi cantly outperform the traditional m ethod t o tune system conguration g lobally ( i.e., g rid s earch) a nd l eads t o h igher e ectiveness th an th e to p pe rforming sy stems of the TREC tracks. We also perform an ablation analysis on the impact of di erent f eatures o n t he model learning capability and show that query expansion features are among the most important for adaptive systems.The study presented in this article is built on the results and conclusions of the previous descriptive analysis studies but moves a step further by performing a predictive analysis: We investigate how system parameters can be set to t a given query, i.e., a query-dependent setting of system parameters. We assume that some parameters of the system can be set on the y at querying time, and a retrieval system allows us to set di erent values for the parameters easily. This is indeed the case for most IR systems nowadays. Retrieval platforms such as Terrier 4 [61], Lemur 5 [70], or Lucene 6 [53] allow us to set parameters for the retrieval step. For example, one may choose between several retrieval models (e.g., BM25, language models), di erent query expansion schemes, and so on. We target this group of parameters that can be set at query time. In contrast, we assume that an IR system has already built an index that cannot be changed easily. For example, it would be di cult to choose between di erent stemmers at query time, unless we construct several indexes using di erent stemmers. We exclude these parameters that cannot be set at query time in this study.The problem we tackle in this article is query-dependent param...

show abstract

Section: Adaptive Systemsmentioning

confidence: 99%

Learning to Adaptively Rank Document Retrieval System Configurations

Deveaud

Mothe

Ullah

et al. 2018

ACM Trans. Inf. Syst.

View full text Add to dashboard Cite

show abstract

“…Before re-programming QECK, we had to create a collection of Q&A pairs. Based on the name of each method-level code snippet in the code corpus, we collected the questions with the "Android, Java" tags and the accepted answer with the "AcceptedAnswer" tag from the stack exchange data dump 17) . This generated Q&A pairs which were indexed as documents consisting of words and SO scores.…”

Section: Qeckmentioning

confidence: 99%

“…To guarantee the quality of the expansion terms, Ref. [17] re-ranks and refines the expansion terms (see "d" in Figure 5). Instead of using a single external expansion source for selecting the expansion term, Ref.…”

Section: Free-form Query Searchmentioning

confidence: 99%

“…(1) We employed Github, stack overflow, and online documentations whereas [16] employed TREC data, Google and Derwent world patents index as the external expansion sources. (2) We did not rank multiple sets of expansion terms whereas [16,17] did this (see a dotted box in red (e) in Figure 5). (3) We produced multiple expansion queries [Q 1 , .…”

Section: Free-form Query Searchmentioning

confidence: 99%

See 1 more Smart Citation

QE-integrating framework based on Github knowledge and SVM ranking

Huang

2019

Sci. China Inf. Sci.

View full text Add to dashboard Cite

The latest query expansion (QE) methods use the software development features for expanding queries. However, these methods allow only one feature to be considered at a time. To consider additional features simultaneously, we propose a QE method based on Github knowledge; this is a new comprehensive feature that covers both the existing features (i.e., the application program interface (API) information and crowd knowledge). It is extracted from the "pull requests" of code repositories on Github, which contain descriptions of a request and its commits, the participants' comments and the API information of the changed files. In addition, we implement a black-box framework that integrates multiple QE methods based on the support vector machine ranking called Github knowledge search repository (GKSR). Our empirical evaluation shows that the GKSR outperforms the state-of-the-art QE methods CodeHow and QECK by 25%-32% in terms of precision.

show abstract

“…In Web search, learning-to-rank (L2R) methods had been used by several researchers [9], [10], [41], [42] to rank the candidate expansion terms. In addition, supervised feature selection approaches based on learning-to-rank algorithm [10], [43] and elastic-net regularization method [11] were employed by researchers to select the effective set of features.…”

Section: Related Workmentioning

confidence: 99%

Query Expansion for Microblog Retrieval Focusing on an Ensemble of Features

Chy

Ullah

Aono

2019

Journal of Information Processing

View full text Add to dashboard Cite

In microblog search, vocabulary mismatch is a persisting problem due to the brevity of tweets and frequent use of unconventional abbreviations. One way of alleviating this problem is to reformulate the query via query expansion. However, finding good expansion terms for a given query is a challenging task. In this paper, we present a query expansion framework, where supervised learning is adopted for selecting expansion terms. Upon retrieving tweets by our proposed topic modeling based query expansion, we utilize the pseudo-relevance feedback and a new temporal relatedness approach to select the candidate tweets. Next, we devise several new features to select the temporally and semantically relevant expansion terms by leveraging the temporal, word embedding, and sentiment association of candidate term and query. Moreover, we also utilize the lexical and twitter specific features to quantify the term relatedness. After supervised feature selection using regularized regression, we estimate the feature importance by applying random forest. Then, we make use of a learning-to-rank (L2R) framework to rank the candidate expansion terms. Results of extensive experiments on TREC Microblog 2011 and 2012 test collections over the Tweets2011 corpus show that our proposed method outperforms the baseline and competitive query expansion methods.

show abstract

Assessment of learning to rank methods for query expansion

Cited by 18 publications

References 30 publications

Learning to Adaptively Rank Document Retrieval System Configurations

Learning to Adaptively Rank Document Retrieval System Configurations

QE-integrating framework based on Github knowledge and SVM ranking

Query Expansion for Microblog Retrieval Focusing on an Ensemble of Features

Contact Info

Product

Resources

About