Ranking is at the core of many artificial intelligence (AI) applications, including search engines, recommender systems, etc. Modern ranking systems are often constructed with learning-to-rank (LTR) models built from user behavior signals. While previous studies have demonstrated the effectiveness of using user behavior signals (e.g., clicks) as both features and labels of LTR algorithms, we argue that existing LTR algorithms that indiscriminately treat behavior and non-behavior signals in input features could lead to suboptimal performance in practice. Because user behavior signals often have strong correlations with the ranking objective and can only be collected on items that have already been shown to users, directly using behavior signals in LTR could create an exploitation bias that hurts the system performance in the long run.

In order to determine the relevance of a given item to a query, most modern search ranking systems make use of features which aggregate prior user behavior for that item and query (e.g. click rate). For practical reasons, when running A/B tests on ranking systems, these features are generally shared between all treatments. For the most common experiment designs, which randomize traffic by user or by session, this creates a pathway by which the behavior of units in one treatment can effect the outcomes for units in other treatments, violating the Stable Unit Treatment Value Assumption (SUTVA) and biasing measured outcomes. Moreover, for experiments targeting improvements to the behavior data available to such features (e.g. online exploration), this pathway is precisely the one we are trying to affect; if such changes occur identically in treatment and control, then they cannot be measured. To address this, we propose the use of experiments which instead randomize traffic based on the search query. To validate our approach, we perform a pair of A/B tests on an explore-exploit framework in the Amazon search page: one under query randomization, and one under user randomization. In line with the theoretical predictions, we find that the query-randomized iteration is able to measure a statistically significant effect (+0.66% Purchases, p=0.001) where the user-randomized iteration does not (-0.02% Purchases, p=0.851).

Cold start is a challenge in product search. Profuse literature addresses related problems such as bias and diversity in search, and cold start is a classic topic in recommender systems research. While search cold start might be seen conceptually as a particular case in such areas, we find that available solutions fail to specifically and practically solve the cold-start problem in product search. The problem is complex as exposing new products may come at the expense of primary business metrics (e.g. revenue), and involves a complex balance between customer satisfaction, seller satisfaction, business performance, short-term gains and long-term value.

Using implicit feedback collected from user clicks as training labels for learning-to-rank algorithms is a well-developed paradigm that has been extensively studied and used in modern IR systems. Using user clicks as ranking features, on the other hand, has not been fully explored in existing literature. Despite its potential in improving short-term system performance, whether the incorporation of user clicks as ranking features is beneficial for learning-to-rank systems in the long term is still questionable. Two of the most important problems are (1) the explicit bias introduced by noisy user behavior, and (2) the implicit bias, which we refer to as the exploitation bias, introduced by the dynamic training and serving of learning-to-rank systems with behavior features. In this paper, we explore the possibility of incorporating user clicks as both training labels and ranking features for learning to rank. We formally investigate the problems in feature collection and model training, and propose a counterfactual feature projection function and a novel uncertainty-aware learning to rank framework. Experiments on public datasets show that ranking models learned with the proposed framework can significantly outperform models built with raw click features and algorithms that rank items without considering model uncertainty.

Seasonality is an important dimension for relevance in e-commerce search. For example, a query jacket has a different set of relevant documents in winter than summer. For an optimal user experience, the e-commerce search engines should incorporate seasonality in product search. In this paper, we formally introduce the concept of seasonal relevance, define it and quantify using data from a major e-commerce store. In our analyses, we find 39% queries are highly seasonally relevant to the time of search and would benefit from handling seasonality in ranking. We propose LogSR and VelSR features to capture product seasonality using state-of-the-art neural models based on self-attention. Comprehensive offline and online experiments over large datasets show the efficacy of our methods to model seasonal relevance. The online A/B test on 784 MM queries shows the treatment with seasonal relevance features results in 2.20% higher purchases and better customer experience overall.

cripts (e.g., Arabic, Greek and Indic languages) one can often find a large amount of user generated transliterated content on the Web in the Roman script. Such content creates a monolingual or cross-lingual space with more than one scripts which is referred as mixed-script space and information retrieval in this space is referred as mixed-script information retrieval (MSIR) [1]. In mixed-script space, the documents and queries may either be in the native script and/or the Roman transliterated script for a language (mono-lingual scenario). There can be further extension of MSIR such as multi-lingual MSIR in which terms can be in multiple scripts in multiple languages. Since there are no standard ways of spelling a word in a non-native script, transliteration content almost always features extensive spelling variations. This phenomenon presents a non-trivial term matching problem for search engines to match the native-script or Roman-transliterated query with the documents in multiple scripts taking into account the spelling variations. This problem, although prevalent inWeb search for users of many languages around the world, has received very little attention till date. Very recently we have formally defined the problem of MSIR and presented the quantitative study on it through Bing query log analysis.

For many languages that use non-Roman based indigenous scripts (e.g., Arabic, Greek and Indic languages) one can often find a large amount of user generated transliterated content on the Web in the Roman script. Such content creates a monolingual or multi-lingual space with more than one script which we refer to as the Mixed-Script space. IR in the mixed-script space is challenging because queries written in either the native or the Roman script need to be matched to the documents written in both the scripts. Moreover, transliterated content features extensive spelling variations. In this paper, we formally introduce the concept of Mixed-Script IR, and through analysis of the query logs of Bing search engine, estimate the prevalence and thereby establish the importance of this problem. We also give a principled solution to handle the mixed-script term matching and spelling variation where the terms across the scripts are modelled jointly in a deep-learning architecture and can be compared in a low-dimensional abstract space. We present an extensive empirical analysis of the proposed method along with the evaluation results in an ad-hoc retrieval setting of mixed-script IR where the proposed method achieves significantly better results (12% increase in MRR and 29% increase in MAP) compared to other state-of-the-art baselines.