While recent pre-trained transformer-based models can perform named entity recognition (NER) with great accuracy, their limited range remains an issue when applied to long documents such as whole novels. To alleviate this issue, a solution is to retrieve relevant context at the document level. Unfortunately, the lack of supervision for such a task means one has to settle for unsupervised approaches. Instead, we propose to generate a synthetic context retrieval training dataset using Alpaca, an instructiontuned large language model (LLM). Using this dataset, we train a neural context retriever based on a BERT model that is able to find relevant context for NER. We show that our method outperforms several retrieval baselines for the NER task on an English literary dataset composed of the first chapter of 40 books.
In the context of public procurement, several indicators called red flags are used to estimate fraud risk. They are computed according to certain contract attributes and are therefore dependent on the proper filling of the contract and award notices. However, these attributes are very often missing in practice, which prohibits red flags computation. Traditional fraud detection approaches focus on tabular data only, considering each contract separately, and are therefore very sensitive to this issue. In this work, we adopt a graph-based method allowing leveraging relations between contracts, to compensate for the missing attributes. We propose PANG (Pattern-Based Anomaly Detection in Graphs), a general supervised framework relying on pattern extraction to detect anomalous graphs in a collection of attributed graphs. Notably, it is able to identify induced subgraphs, a type of pattern widely overlooked in the literature. When benchmarked on standard datasets, its predictive performance is on par with state-of-the-art methods, with the additional advantage of being explainable. These experiments also reveal that induced patterns are more discriminative on certain datasets. When applying PANG to public procurement data, the prediction is superior to other methods, and it identifies subgraph patterns that are characteristic of fraud-prone situations, thereby making it possible to better understand fraudulent behavior.
Pre-trained transformer-based models have recently shown great performance when applied to Named Entity Recognition (NER). As the complexity of their self-attention mechanism prevents them from processing long documents at once, these models are usually applied in a sequential fashion. Such an approach unfortunately only incorporates local context and prevents leveraging global document context in long documents such as novels, which might hinder performance. In this article, we explore the impact of global document context, and its relationships with local context. We find that correctly retrieving global document context has a greater impact on performance than only leveraging local context, prompting for further research on how to better retrieve that context.
Named Entity Recognition (NER) is a low-level task often used as a foundation for solving higher level NLP problems. In the context of character detection in novels, NER false negatives can be an issue as they possibly imply missing certain characters or relationships completely. In this article, we demonstrate that applying a straightforward data augmentation technique allows training a model achieving higher recall, at the cost of a certain amount of precision regarding ambiguous entities. We show that this decrease in precision can be mitigated by giving the model more local context, which resolves some of the ambiguities.
In the context of cluster analysis and graph partitioning, many external evaluation measures have been proposed in the literature to compare two partitions of the same set. This makes the task of selecting the most appropriate measure for a given situation a challenge for the end user. However, this issue is overlooked in the literature. Researchers tend to follow tradition and use the standard measures of their field, although they often became standard only because previous researchers started consistently using them. In this work, we propose a new empirical evaluation framework to solve this issue, and help the end user selecting an appropriate measure for their application. For a collection of candidate measures, it first consists in describing their behavior by computing them for a generated dataset of partitions, obtained by applying a set of predefined parametric partition transformations. Second, our framework performs a regression analysis to characterize the measures in terms of how they are affected by these parameters and transformations. This allows both describing and comparing the measures. Our approach is not tied to any specific measure or application, so it can be applied to any situation. We illustrate its relevance by applying it to a selection of standard measures, and show how it can be put in practice through two concrete use cases.
In order to study real-world systems, many applied works model them through signed graphs, i.e. graphs whose edges are labeled as either positive or negative. Such a graph is considered as structurally balanced when it can be partitioned into a number of modules, such that positive (resp. negative) edges are located inside (resp. in-between) the modules. When it is not the case, authors look for the closest partition to such balance, a problem called Correlation Clustering (CC). Due to the complexity of the CC problem, the standard approach is to find a single optimal partition and stick to it, even if other optimal or high scoring solutions possibly exist. In this work, we study the space of optimal solutions of the CC problem, on a collection of synthetic complete graphs. We show empirically that under certain conditions, there can be many optimal partitions of a signed graph. Some of these are very different and thus provide distinct perspectives on the system, as illustrated on a small real-world graph. This is an important result, as it implies that one may have to find several, if not all, optimal solutions of the CC problem, in order to properly study the considered system.
With the spread of online social networks, it is more and more difficult to monitor all the user-generated content. Automating the moderation process of the inappropriate exchange content on Internet has thus become a priority task. Methods have been proposed for this purpose, but it can be challenging to find a suitable dataset to train and develop them. This issue is especially true for approaches based on information derived from the structure and the dynamic of the conversation. In this work, we propose an original framework, based on the Wikipedia Comment corpus, with comment-level abuse annotations of different types. The major contribution concerns the reconstruction of conversations, by comparison to existing corpora, which focus only on isolated messages (i.e. taken out of their conversational context). This large corpus of more than 380k annotated messages opens perspectives for online abuse detection and especially for context-based approaches. We also propose, in addition to this corpus, a complete benchmarking platform to stimulate and fairly compare scientific works around the problem of content abuse detection, trying to avoid the recurring problem of result replication. Finally, we apply two classification methods to our dataset to demonstrate its potential.
A character network is a graph extracted from a narrative, in which vertices represent characters and edges correspond to interactions between them. A number of narrative-related problems can be addressed automatically through the analysis of character networks, such as summarization, classification, or role detection. Character networks are particularly relevant when considering works of fictions (e.g. novels, plays, movies, TV series), as their exploitation allows developing information retrieval and recommendation systems. However, works of fiction possess specific properties making these tasks harder. This survey aims at presenting and organizing the scientific literature related to the extraction of character networks from works of fiction, as well as their analysis. We first describe the extraction process in a generic way, and explain how its constituting steps are implemented in practice, depending on the medium of the narrative, the goal of the network analysis, and other factors. We then review the descriptive tools used to characterize character networks, with a focus on the way they are interpreted in this context. We illustrate the relevance of character networks by also providing a review of applications derived from their analysis. Finally, we identify the limitations of the existing approaches, and the most promising perspectives.
In recent years, online social networks have allowed worldwide users to meet and discuss. As guarantors of these communities, the administrators of these platforms must prevent users from adopting inappropriate behaviors. This verification task, mainly done by humans, is more and more difficult due to the ever growing amount of messages to check. Methods have been proposed to automatize this moderation process, mainly by providing approaches based on the textual content of the exchanged messages. Recent work has also shown that characteristics derived from the structure of conversations, in the form of conversational graphs, can help detecting these abusive messages. In this paper, we propose to take advantage of both sources of information by proposing fusion methods integrating content-and graph-based features. Our experiments on raw chat logs show that the content of the messages, but also of their dynamics within a conversation contain partially complementary information, allowing performance improvements on an abusive message classification task with a final F-measure of 93.26%.
Moderation of user-generated content in an online community is a challenge that has great socio-economical ramifications. However, the costs incurred by delegating this work to human agents are high. For this reason, an automatic system able to detect abuse in user-generated content is of great interest. There are a number of ways to tackle this problem, but the most commonly seen in practice are word filtering or regular expression matching. The main limitations are their vulnerability to intentional obfuscation on the part of the users, and their context-insensitive nature. Moreover, they are language-dependent and may require appropriate corpora for training. In this paper, we propose a system for automatic abuse detection that completely disregards message content. We first extract a conversational network from raw chat logs and characterize it through topological measures. We then use these as features to train a classifier on our abuse detection task. We thoroughly assess our system on a dataset of user comments originating from a French Massively Multiplayer Online Game. We identify the most appropriate network extraction parameters and discuss the discriminative power of our features, relatively to their topological and temporal nature. Our method reaches an F-measure of 83.89 when using the full feature set, improving on existing approaches. With a selection of the most discriminative features, we dramatically cut computing time while retaining most of the performance (82.65).