How to Do Things With Words

National Institute of Standards and Technology における超伝導研究及び生活

This tutorial is concerned with a comprehensive introduction to the research area of learning to rank for information retrieval. In the first part of the tutorial, we will introduce three major approaches to learning to rank, i.e., the pointwise, pairwise, and listwise approaches, analyze the relationship between the loss functions used in these approaches and the widely-used IR evaluation measures, evaluate the performance of these approaches on the LETOR benchmark datasets, and demonstrate how to use these approaches to solve real ranking applications. In the second part of the tutorial, we will discuss some advanced topics regarding learning to rank, such as relational ranking, diverse ranking, semi-supervised ranking, transfer ranking, query-dependent ranking, and training data preprocessing. In the third part, we will briefly mention the recent advances on statistical learning theory for ranking, which explain the generalization ability and statistical consistency of different ranking methods. In the last part, we will conclude the tutorial and show several future research directions.

/pdf/learning-to-rank-for-information-retrieval-2q6t3viyug.pdf

Learning to Rank for Information Retrieval

Today is the era of the Internet of Things (IoT). The recent advances in hardware and information technology have accelerated the deployment of billions of interconnected, smart and adaptive devices in critical infrastructures like health, transportation, environmental control, and home automation. Transferring data over a network without requiring any kind of human-to-computer or human-to-human interaction, brings reliability and convenience to consumers, but also opens a new world of opportunity for intruders, and introduces a whole set of unique and complicated questions to the field of Digital Forensics. Although IoT data could be a rich source of evidence, forensics professionals cope with diverse problems, starting from the huge variety of IoT devices and non-standard formats, to the multi-tenant cloud infrastructure and the resulting multi-jurisdictional litigations. A further challenge is the end-to-end encryption which represents a trade-off between users’ right to privacy and the success of the forensics investigation. Due to its volatile nature, digital evidence has to be acquired and analyzed using validated tools and techniques that ensure the maintenance of the Chain of Custody. Therefore, the purpose of this paper is to identify and discuss the main issues involved in the complex process of IoT-based investigations, particularly all legal, privacy and cloud security challenges. Furthermore, this work provides an overview of the past and current theoretical models in the digital forensics science. Special attention is paid to frameworks that aim to extract data in a privacy-preserving manner or secure the evidence integrity using decentralized blockchain-based solutions. In addition, the present paper addresses the ongoing Forensics-as-a-Service (FaaS) paradigm, as well as some promising cross-cutting data reduction and forensics intelligence techniques. Finally, several other research trends and open issues are presented, with emphasis on the need for proactive Forensics Readiness strategies and generally agreed-upon standards.

/pdf/a-survey-on-the-internet-of-things-iot-forensics-challenges-3w6jzejgc2.pdf

A Survey on the Internet of Things (IoT) Forensics: Challenges, Approaches, and Open Issues

A Systematic Review of the Smart Home Literature: A user perspective

Data fusion is the combination of the results of independent searches on a document collection into one single output result set. It has been shown in the past that this can greatly improve retrieval effectiveness over that of the individual results.This paper presents probFuse, a probabilistic approach to data fusion. ProbFuse assumes that the performance of the individual input systems on a number of training queries is indicative of their future performance. The fused result set is based on probabilities of relevance calculated during this training process. Retrieval experiments using data from the TREC ad hoc collection demonstrate that probFuse achieves results superior to that of the popular CombMNZ fusion algorithm.

ProbFuse: a probabilistic approach to data fusion

Given the ever-increasing prevalence of technology in modern life, there is a corresponding increase in the likelihood of digital devices being pertinent to a criminal investigation or civil litigation. As a direct consequence, the number of investigations requiring digital forensic expertise is resulting in huge digital evidence backlogs being encountered by law enforcement agencies throughout the world. It can be anticipated that the number of cases requiring digital forensic analysis will greatly increase in the future. It is also likely that each case will require the analysis of an increasing number of devices including computers, smartphones, tablets, cloud-based services, Internet of Things devices, wearables, etc. The variety of new digital evidence sources pose new and challenging problems for the digital investigator from an identification, acquisition, storage and analysis perspective. This paper explores the current challenges contributing to the backlog in digital forensics from a technical standpoint and outlines a number of future research topics that could greatly contribute to a more efficient digital forensic process.

Current Challenges and Future Research Areas for Digital Forensic Investigation

Word embeddings act as an important component of deep models for providing input features in downstream language tasks, such as sequence labelling and text classification. In the last decade, a substantial number of word embedding methods have been proposed for this purpose, mainly falling into the categories of classic and context-based word embeddings. In this paper, we conduct controlled experiments to systematically examine both classic and contextualised word embeddings for the purposes of text classification. To encode a sequence from word representations, we apply two encoders, namely CNN and BiLSTM, in the downstream network architecture. To study the impact of word embeddings on different datasets, we select four benchmarking classification datasets with varying average sample length, comprising both single-label and multi-label classification tasks. The evaluation results with confidence intervals indicate that CNN as the downstream encoder outperforms BiLSTM in most situations, especially for document context-insensitive datasets. This study recommends choosing CNN over BiLSTM for document classification datasets where the context in sequence is not as indicative of class membership as sentence datasets. For word embeddings, concatenation of multiple classic embeddings or increasing their size does not lead to a statistically significant difference in performance despite a slight improvement in some cases. For context-based embeddings, we studied both ELMo and BERT. The results show that BERT overall outperforms ELMo, especially for long document datasets. Compared with classic embeddings, both achieve an improved performance for short datasets while the improvement is not observed in longer datasets.

https://dl.acm.org/doi/pdf/10.1145/3443279.3443304

A Comparative Study on Word Embeddings in Deep Learning for Text Classification

11th Annual ADFSL Conference on Digital Forensics, Security and Law (CDFSL 2016), Daytona Beach, Florida, USA, 24-26 May 2016

/pdf/current-challenges-and-future-research-areas-for-digital-31b262hzzf.pdf

Data Fusion is the combination of a number of independent search results, relating to the same document collection, into a single result to be presented to the user. A number of probabilistic data fusion models have been shown to be effective in empirical studies. These typically attempt to estimate the probability that particular documents will be relevant, based on training data. However, little attempt has been made to gauge how the accuracy of these estimations affect fusion performance. The focus of this paper is twofold: firstly, that accurate estimation of the probability of relevance results in effective data fusion; and secondly, that an effective approximation of this probability can be made based on less training data that has previously been employed. This is based on the observation that the distribution of relevant documents follows a similar pattern in most high-quality result sets. Curve fitting suggests that this can be modelled by a simple function that is less complex than other models that have been proposed. The use of existing IR evaluation metrics is proposed as a substitution for probability calculations. Mean Average Precision is used to demonstrate the effectiveness of this approach, with evaluation results demonstrating competitive performance when compared with related algorithms with more onerous requirements for training data.

David Lillis

Papers

ProbFuse: a probabilistic approach to data fusion

Current Challenges and Future Research Areas for Digital Forensic Investigation

A Comparative Study on Word Embeddings in Deep Learning for Text Classification

Current Challenges and Future Research Areas for Digital Forensic Investigation

Estimating probabilities for effective data fusion