Papers by Jean-Michel Renders
Proceedings of the 1st Workshop on Representation Learning for NLP, 2016
We consider the problem of computing a sequence of rankings that maximizes consumer-side utility ... more We consider the problem of computing a sequence of rankings that maximizes consumer-side utility while minimizing producer-side individual unfairness of exposure. While prior work has addressed this problem using linear or quadratic programs on bistochastic matrices, such approaches, relying on Birkhoff-von Neumann (BvN) decompositions, are too slow to be implemented at large scale. In this paper we introduce a geometrical object, a polytope that we call expohedron, whose points represent all achievable exposures of items for a Position Based Model (PBM). We exhibit some of its properties and lay out a Carathéodory decomposition algorithm with complexity O(n^2log(n)) able to express any point inside the expohedron as a convex sum of at most n vertices, where n is the number of items to rank. Such a decomposition makes it possible to express any feasible target exposure as a distribution over at most n rankings. Furthermore we show that we can use this polytope to recover the whole P...
Proceedings of the 22nd International Conference on World Wide Web - WWW '13 Companion, 2013
Lecture Notes in Computer Science, 2020
Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, 2022
(a) = 2. A line segment in R 2 (b) = 3. A polygon in R 3 (c) = 4. A polyhedron in R 4 Figure 1: E... more (a) = 2. A line segment in R 2 (b) = 3. A polygon in R 3 (c) = 4. A polyhedron in R 4 Figure 1: Examples of the DCG expohedron for ∈ {2, 3, 4} items. The vertices are the DCG exposures 1 log 2 (2) ,. .. , 1 log 2 (+1) under application of the symmetric group S. The expohedron is the convex hull of these vertices. Expohedra live in hyperplanes of dimension − 1.
Spoken dialogue systems typically use a list of top-N ASR hypotheses for inferring the semantic m... more Spoken dialogue systems typically use a list of top-N ASR hypotheses for inferring the semantic meaning and tracking the state of the dialogue. However ASR graphs, such as confusion networks (confnets), provide a compact representation of a richer hypothesis space than a top-N ASR list. In this paper, we study the benefits of using confusion networks with a state-of-the-art neural dialogue state tracker (DST). We encode the 2-dimensional confnet into a 1-dimensional sequence of embeddings using an attentional confusion network encoder which can be used with any DST system. Our confnet encoder is plugged into the state-of-the-art 'Global-locally Self-Attentive Dialogue State Tacker' (GLAD) model for DST and obtains significant improvements in both accuracy and inference time compared to using top-N ASR hypotheses.
Proceedings of the First IEEE Conference on Evolutionary Computation. IEEE World Congress on Computational Intelligence
Abstract Two methods of hybridizing genetic algorithms (GA) with hill-climbing for global optimiz... more Abstract Two methods of hybridizing genetic algorithms (GA) with hill-climbing for global optimization are investigated. The first one involves two interwoven levels of optimization-evolution (GA) and individual learning (hill-climbing)-which cooperate in the global ...
Abstract: News and social media are emerging as a dominant source of information for numerous app... more Abstract: News and social media are emerging as a dominant source of information for numerous applications. However, their vast unstructured content present challenges to efficient extraction of such information. In this paper, we present the SYNC3 system that aims to intelligently structure content from both traditional news media and the blogosphere. To achieve this goal, SYNC3 incorporates innovative algorithms that first model news media content statistically, based on fine clustering of articles into so-called “news events”. Such models are then adapted and applied to the blogosphere domain, allowing its content to map to the traditional news domain. Furthermore, appropriate algorithms are employed to extract news event labels and relations between events, in order to efficiently present news content to the system end users. 1.
The aim of this document is to describe the methods we used in the Patent Image
MACHINE learning techniques have been used for various tasks of document management and textual i... more MACHINE learning techniques have been used for various tasks of document management and textual information access, such as categorisation, information extraction, or automatic organization of large document collections.
Identifying potentially responsive data using keyword searches (and their variants) has become st... more Identifying potentially responsive data using keyword searches (and their variants) has become standard operating procedure in large-scale document reviews in litigation, regulatory inquiries and subpoena compliance. At the same time, there is growing skepticism within the legal community as to the adequacy of such an approach. Developments in information retrieval and extraction technologies have lead to a number of more sophisticated approaches to meeting the challenges of isolating responsive material in complex litigation and regulatory matters. Initially met with resistance by judges, practicing attorneys and legal professionals, such approaches are garnering more serious analysis as the amount of potential source data expands and the costs of collection, processing and most significantly, review, strain corporate budgets. One of these new approaches, applying machine learning classification to the human decision making process in litigation document reviews is the subject of t...
Lecture Notes in Computer Science, 2018
We consider the problem of Active Search, where a maximum of relevant objects-ideally all relevan... more We consider the problem of Active Search, where a maximum of relevant objects-ideally all relevant objects-should be retrieved with the minimum effort or minimum time. Typically, there are two main challenges to face when tackling this problem: first, the class of relevant objects has often low prevalence and, secondly, this class can be multifaceted or multi-modal: objects could be relevant for completely different reasons. To solve this problem and its associated issues, we propose an approach based on a non-stationary (aka restless) extension of Thompson Sampling, a well-known strategy for Multi-Armed Bandits problems. The collection is first soft-clustered into a finite set of components and a posterior distribution of getting a relevant object inside each cluster is updated after receiving the user feedback about the proposed instances. The "next instance" selection strategy is a mixed, two-level decision process, where both the soft clusters and their instances are considered. This method can be considered as an insurance, where the cost of the insurance is an extra exploration effort in the short run, for achieving a nearly "total" recall with less efforts in the long run.
Proceedings of the 1992 IEEE International Symposium on Intelligent Control
ABSTRACT
ArXiv, 2015
We propose an approach for helping agents compose email replies to customer requests. To enable t... more We propose an approach for helping agents compose email replies to customer requests. To enable that, we use LDA to extract latent topics from a collection of email exchanges. We then use these latent topics to label our data, obtaining a so-called "silver standard" topic labelling. We exploit this labelled set to train a classifier to: (i) predict the topic distribution of the entire agent's email response, based on features of the customer's email; and (ii) predict the topic distribution of the next sentence in the agent's reply, based on the customer's email features and on features of the agent's current sentence. The experimental results on a large email collection from a contact center in the tele- com domain show that the proposed ap- proach is effective in predicting the best topic of the agent's next sentence. In 80% of the cases, the correct topic is present among the top five recommended topics (out of fifty possible ones). This shows the...
ArXiv, 2017
We consider the optimal value of information (VoI) problem, where the goal is to sequentially sel... more We consider the optimal value of information (VoI) problem, where the goal is to sequentially select a set of tests with a minimal cost, so that one can efficiently make the best decision based on the observed outcomes. Existing algorithms are either heuristics with no guarantees, or scale poorly (with exponential run time in terms of the number of available tests). Moreover, these methods assume a known distribution over the test outcomes, which is often not the case in practice. We propose an efficient sampling-based online learning framework to address the above issues. First, assuming the distribution over hypotheses is known, we propose a dynamic hypothesis enumeration strategy, which allows efficient information gathering with strong theoretical guarantees. We show that with sufficient amount of samples, one can identify a near-optimal decision with high probability. Second, when the parameters of the hypotheses distribution are unknown, we propose an algorithm which learns th...
ArXiv, 2021
Information retrieval (IR) systems traditionally aim to maximize metrics built on rankings, such ... more Information retrieval (IR) systems traditionally aim to maximize metrics built on rankings, such as precision or NDCG. However, the non-differentiability of the ranking operation prevents direct optimization of such metrics in state-of-the-art neural IR models, which rely entirely on the ability to compute meaningful gradients. To address this shortcoming, we propose SmoothI, a smooth approximation of rank indicators that serves as a basic building block to devise differentiable approximations of IR metrics. We further provide theoretical guarantees on SmoothI and derived approximations, showing in particular that the approximation errors decrease exponentially with an inverse temperature-like hyperparameter that controls the quality of the approximations. Extensive experiments conducted on four standard learning-to-rank datasets validate the efficacy of the listwise losses based on SmoothI, in comparison to previously proposed ones. Additional experiments with a vanilla BERT rankin...
Spoken dialogue systems typically use a list of top-N ASR hypotheses for inferring the semantic m... more Spoken dialogue systems typically use a list of top-N ASR hypotheses for inferring the semantic meaning and tracking the state of the dialogue. However ASR graphs, such as confusion networks (confnets), provide a compact representation of a richer hypothesis space than a top-N ASR list. In this paper, we study the benefits of using confusion networks with a state-of-the-art neural dialogue state tracker (DST). We encode the 2-dimensional confnet into a 1-dimensional sequence of embeddings using an attentional confusion network encoder which can be used with any DST system. Our confnet encoder is plugged into the state-of-the-art 'Global-locally Self-Attentive Dialogue State Tacker' (GLAD) model for DST and obtains significant improvements in both accuracy and inference time compared to using top-N ASR hypotheses.
Most spoken dialogue systems decide whether to accept or reject results from the speech recogniti... more Most spoken dialogue systems decide whether to accept or reject results from the speech recognition component by applying a threshold to the associated confidence score. We show how a simple and general method, based on standard approaches to document classification using Support Vector Machines, can give substantially better performance. Experiments carried out on a medium-vocabulary command and control task show a relative reduction in the task-level error rate by about 25% compared to the baseline confidence threshold method.
In our participation to the TREC 2019 Fair Ranking Track, the University of Glasgow Terrier Team ... more In our participation to the TREC 2019 Fair Ranking Track, the University of Glasgow Terrier Team and Naver Labs Europe joined forces to investigate (1) a novel probabilistic retrieval strategy that maximises the utility of the ranking, (2) two greedy brute-force reranking approaches that build on our novel probabilistic retrieval strategy and enforce individual fairness before adopting a particular trade-off between the utility and the fairness of the ranking, and (3) two approaches that deploy search results diversification as a fairness component to diversify over multiple possible dimensions of the task’s unknown author groupings.
Uploads
Papers by Jean-Michel Renders