arXiv subjects

Ansgar Scherp

Publications and source records attributed to Ansgar Scherp.

54 records · Page 3Linked to original sources

arXiv2022

Analysis of GraphSum's Attention Weights to Improve the Explainability of Multi-Document Summarization

Modern multi-document summarization (MDS) methods are based on transformer architectures. They generate state of the art summaries, but lack explainability. We focus on graph-based transformer models for MDS as they gained recent popularity. We aim to improve the explainability of the graph-based MDS by analyzing their attention weights. In a graph-based MDS such as GraphSum, vertices represent the textual units, while the edges form some similarity graph over the units. We compare GraphSum's performance utilizing different textual units, i. e., sentences versus paragraphs, on two news benchmark datasets, namely WikiSum and MultiNews. Our experiments show that paragraph-level representations provide the best summarization performance. Thus, we subsequently focus oAnalysisn analyzing the paragraph-level attention weights of GraphSum's multi-heads and decoding layers in order to improve the explainability of a transformer-based MDS model. As a reference metric, we calculate the ROUGE scores between the input paragraphs and each sentence in the generated summary, which indicate source origin information via text similarity. We observe a high correlation between the attention weights and this reference metric, especially on the the later decoding layers of the transformer architecture. Finally, we investigate if the generated summaries follow a pattern of positional bias by extracting which paragraph provided the most information for each generated summary. Our results show that there is a high correlation between the position in the summary and the source origin.

cs.CL↗

arXiv2022

Time and Memory Efficient Parallel Algorithm for Structural Graph Summaries and two Extensions to Incremental Summarization and $k$-Bisimulation for Long $k$-Chaining

We developed a flexible parallel algorithm for graph summarization based on vertex-centric programming and parameterized message passing. The base algorithm supports infinitely many structural graph summary models defined in a formal language. An extension of the parallel base algorithm allows incremental graph summarization. In this paper, we prove that the incremental algorithm is correct and show that updates are performed in time $\mathcal{O}(Δ\cdot d^k)$, where $Δ$ is the number of additions, deletions, and modifications to the input graph, $d$ the maximum degree, and $k$ is the maximum distance in the subgraphs considered. Although the iterative algorithm supports values of $k>1$, it requires nested data structures for the message passing that are memory-inefficient. Thus, we extended the base summarization algorithm by a hash-based messaging mechanism to support a scalable iterative computation of graph summarizations based on $k$-bisimulation for arbitrary $k$. We empirically evaluate the performance of our algorithms using benchmark and real-world datasets. The incremental algorithm almost always outperforms the batch computation. We observe in our experiments that the incremental algorithm is faster even in cases when $50\%$ of the graph database changes from one version to the next. The incremental computation requires a three-layered hash index, which has a low memory overhead of only $8\%$ ($\pm 1\%$). Finally, the incremental summarization algorithm outperforms the batch algorithm even with fewer cores. The iterative parallel $k$-bisimulation algorithm computes summaries on graphs with over $10$M edges within seconds. We show that the algorithm processes graphs of $100+\,$M edges within a few minutes while having a moderate memory consumption of $<150$ GB. For the largest BSBM1B dataset with 1 billion edges, it computes $k=10$ bisimulation in under an hour.

cs.DS↗

arXiv2022

General Cross-Architecture Distillation of Pretrained Language Models into Matrix Embeddings

Large pretrained language models (PreLMs) are revolutionizing natural language processing across all benchmarks. However, their sheer size is prohibitive for small laboratories or for deployment on mobile devices. Approaches like pruning and distillation reduce the model size but typically retain the same model architecture. In contrast, we explore distilling PreLMs into a different, more efficient architecture, Continual Multiplication of Words (CMOW), which embeds each word as a matrix and uses matrix multiplication to encode sequences. We extend the CMOW architecture and its CMOW/CBOW-Hybrid variant with a bidirectional component for more expressive power, per-token representations for a general (task-agnostic) distillation during pretraining, and a two-sequence encoding scheme that facilitates downstream tasks on sentence pairs, such as sentence similarity and natural language inference. Our matrix-based bidirectional CMOW/CBOW-Hybrid model is competitive to DistilBERT on question similarity and recognizing textual entailment, but uses only half of the number of parameters and is three times faster in terms of inference speed. We match or exceed the scores of ELMo for all tasks of the GLUE benchmark except for the sentiment analysis task SST-2 and the linguistic acceptability task CoLA. However, compared to previous cross-architecture distillation approaches, we demonstrate a doubling of the scores on detecting linguistic acceptability. This shows that matrix-based embeddings can be used to distill large PreLM into competitive models and motivates further research in this direction.

cs.CL↗

arXiv2022

Bag-of-Words vs. Graph vs. Sequence in Text Classification: Questioning the Necessity of Text-Graphs and the Surprising Strength of a Wide MLP

Graph neural networks have triggered a resurgence of graph-based text classification methods, defining today's state of the art. We show that a wide multi-layer perceptron (MLP) using a Bag-of-Words (BoW) outperforms the recent graph-based models TextGCN and HeteGCN in an inductive text classification setting and is comparable with HyperGAT. Moreover, we fine-tune a sequence-based BERT and a lightweight DistilBERT model, which both outperform all state-of-the-art models. These results question the importance of synthetic graphs used in modern text classifiers. In terms of efficiency, DistilBERT is still twice as large as our BoW-based wide MLP, while graph-based models like TextGCN require setting up an $\mathcal{O}(N^2)$ graph, where $N$ is the vocabulary plus corpus size. Finally, since Transformers need to compute $\mathcal{O}(L^2)$ attention weights with sequence length $L$, the MLP models show higher training and inference speeds on datasets with long sequences.

cs.CL↗

arXiv2021

Lifelong Learning of Graph Neural Networks for Open-World Node Classification

Graph neural networks (GNNs) have emerged as the standard method for numerous tasks on graph-structured data such as node classification. However, real-world graphs are often evolving over time and even new classes may arise. We model these challenges as an instance of lifelong learning, in which a learner faces a sequence of tasks and may take over knowledge acquired in past tasks. Such knowledge may be stored explicitly as historic data or implicitly within model parameters. In this work, we systematically analyze the influence of implicit and explicit knowledge. Therefore, we present an incremental training method for lifelong learning on graphs and introduce a new measure based on $k$-neighborhood time differences to address variances in the historic data. We apply our training method to five representative GNN architectures and evaluate them on three new lifelong node classification datasets. Our results show that no more than 50% of the GNN's receptive field is necessary to retain at least 95% accuracy compared to training over the complete history of the graph data. Furthermore, our experiments confirm that implicit knowledge becomes more important when fewer explicit knowledge is available.

cs.LG↗

arXiv2021

Recommendations for Item Set Completion: On the Semantics of Item Co-Occurrence With Data Sparsity, Input Size, and Input Modalities

We address the problem of recommending relevant items to a user in order to "complete" a partial set of items already known. We consider the two scenarios of citation and subject label recommendation, which resemble different semantics of item co-occurrence: relatedness for co-citations and diversity for subject labels. We assess the influence of the completeness of an already known partial item set on the recommender performance. We also investigate data sparsity through a pruning parameter and the influence of using additional metadata. As recommender models, we focus on different autoencoders, which are particularly suited for reconstructing missing items in a set. We extend autoencoders to exploit a multi-modal input of text and structured data. Our experiments on six real-world datasets show that supplying the partial item set as input is helpful when item co-occurrence resembles relatedness, while metadata are effective when co-occurrence implies diversity. This outcome means that the semantics of item co-occurrence is an important factor. The simple item co-occurrence model is a strong baseline for citation recommendation. However, autoencoders have the advantage to enable exploiting additional metadata besides the partial item set as input and achieve comparable performance. For the subject label recommendation task, the title is the most important attribute. Adding more input modalities sometimes even harms the result. In conclusion, it is crucial to consider the semantics of the item co-occurrence for the choice of an appropriate recommendation model and carefully decide which metadata to exploit.

cs.IR↗

arXiv2021

A Comparison of Deep-Learning Methods for Analysing and Predicting Business Processes

Deep-learning models such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) have been successfully used for process-mining tasks. They have achieved better performance for different predictive tasks than traditional approaches. We extend the existing body of research by testing four different variants of Graph Neural Networks (GNN) and a fully connected Multi-layer Perceptron (MLP) with dropout for the tasks of predicting the nature and timestamp of the next process activity. In contrast to existing studies, we evaluate our models' performance at different stages of a process, determined by quartiles of the number of events and normalized quarters of the case duration. This provides new insights into the performance of a prediction model, as they behave differently at different stages of a business-process. Interestingly, our experiments show that the simple MLP often outperforms more sophisticated deep-learning models in both prediction tasks. We argue that care needs to be taken when applying automated process-prediction techniques at different stages of a process. We further argue that researchers should reflect their results with strong baselines methods like simple MLPs.

cs.LG↗

arXiv2021

FLUID: A Common Model for Semantic Structural Graph Summaries Based on Equivalence Relations

Summarization is a widespread method for handling very large graphs. The task of structural graph summarization is to compute a concise but meaningful synopsis of the key structural information of a graph. As summaries may be used for many different purposes, there is no single concept or model of graph summaries. We have studied existing structural graph summaries for large-scale (semantic) graphs. Despite their different concepts and purposes, we found commonalities in the graph structures they capture. We use these commonalities to provide for the first time a formally defined common model, FLUID (FLexible graph sUmmarIes for Data graphs), that allows us to flexibly define structural graph summaries. FLUID allows graph summaries to be quickly defined, adapted, and compared for different purposes and datasets. To this end, FLUID provides features of structural summarization based on equivalence relations such as distinction of types and properties, direction of edges, bisimulation, and inference. We conduct a detailed complexity analysis of the features provided by FLUID. We show that graph summaries defined with FLUID can be computed in the worst case in time $\mathcal{O}(n^2)$ w.r.t. $n$, the number of edges in the data graph. An empirical analysis of large-scale web graphs with billions of edges indicates a typical running time of $Θ(n)$. Based on the formal FLUID model, one can quickly define and modify various structural graph summaries from the literature and beyond.

cs.DB↗

arXiv2020

Indexing Data on the Web: A Comparison of Schema-level Indices for Data Search -- Extended Technical Report

Indexing the Web of Data offers many opportunities, in particular, to find and explore data sources. One major design decision when indexing the Web of Data is to find a suitable index model, i.e., how to index and summarize data. Various efforts have been conducted to develop specific index models for a given task. With each index model designed, implemented, and evaluated independently, it remains difficult to judge whether an approach generalizes well to another task, set of queries, or dataset. In this work, we empirically evaluate six representative index models with unique feature combinations. Among them is a new index model incorporating inferencing over RDFS and owl:sameAs. We implement all index models for the first time into a single, stream-based framework. We evaluate variations of the index models considering sub-graphs of size 0, 1, and 2 hops on two large, real-world datasets. We evaluate the quality of the indices regarding the compression ratio, summarization ratio, and F1-score denoting the approximation quality of the stream-based index computation. The experiments reveal huge variations in compression ratio, summarization ratio, and approximation quality for different index models, queries, and datasets. However, we observe meaningful correlations in the results that help to determine the right index model for a given task, type of query, and dataset.

cs.DB↗

arXiv2019

Multi-Modal Adversarial Autoencoders for Recommendations of Citations and Subject Labels

We present multi-modal adversarial autoencoders for recommendation and evaluate them on two different tasks: citation recommendation and subject label recommendation. We analyze the effects of adversarial regularization, sparsity, and different input modalities. By conducting 408 experiments, we show that adversarial regularization consistently improves the performance of autoencoders for recommendation. We demonstrate, however, that the two tasks differ in the semantics of item co-occurrence in the sense that item co-occurrence resembles relatedness in case of citations, yet implies diversity in case of subject labels. Our results reveal that supplying the partial item set as input is only helpful, when item co-occurrence resembles relatedness. When facing a new recommendation task it is therefore crucial to consider the semantics of item co-occurrence for the choice of an appropriate model.

cs.IR↗

arXiv2019

Can Graph Neural Networks Go "Online"? An Analysis of Pretraining and Inference

Large-scale graph data in real-world applications is often not static but dynamic, i. e., new nodes and edges appear over time. Current graph convolution approaches are promising, especially, when all the graph's nodes and edges are available during training. When unseen nodes and edges are inserted after training, it is not yet evaluated whether up-training or re-training from scratch is preferable. We construct an experimental setup, in which we insert previously unseen nodes and edges after training and conduct a limited amount of inference epochs. In this setup, we compare adapting pretrained graph neural networks against retraining from scratch. Our results show that pretrained models yield high accuracy scores on the unseen nodes and that pretraining is preferable over retraining from scratch. Our experiments represent a first step to evaluate and develop truly online variants of graph neural networks.

cs.LG↗

arXiv2019

CBOW Is Not All You Need: Combining CBOW with the Compositional Matrix Space Model

Continuous Bag of Words (CBOW) is a powerful text embedding method. Due to its strong capabilities to encode word content, CBOW embeddings perform well on a wide range of downstream tasks while being efficient to compute. However, CBOW is not capable of capturing the word order. The reason is that the computation of CBOW's word embeddings is commutative, i.e., embeddings of XYZ and ZYX are the same. In order to address this shortcoming, we propose a learning algorithm for the Continuous Matrix Space Model, which we call Continual Multiplication of Words (CMOW). Our algorithm is an adaptation of word2vec, so that it can be trained on large quantities of unlabeled text. We empirically show that CMOW better captures linguistic properties, but it is inferior to CBOW in memorizing word content. Motivated by these findings, we propose a hybrid model that combines the strengths of CBOW and CMOW. Our results show that the hybrid CBOW-CMOW-model retains CBOW's strong ability to memorize word content while at the same time substantially improving its ability to encode other linguistic information by 8%. As a result, the hybrid also performs better on 8 out of 11 supervised downstream tasks with an average improvement of 1.2%.

cs.CL↗

arXiv2018

Using Deep Learning for Title-Based Semantic Subject Indexing to Reach Competitive Performance to Full-Text

For (semi-)automated subject indexing systems in digital libraries, it is often more practical to use metadata such as the title of a publication instead of the full-text or the abstract. Therefore, it is desirable to have good text mining and text classification algorithms that operate well already on the title of a publication. So far, the classification performance on titles is not competitive with the performance on the full-texts if the same number of training samples is used for training. However, it is much easier to obtain title data in large quantities and to use it for training than full-text data. In this paper, we investigate the question how models obtained from training on increasing amounts of title training data compare to models from training on a constant number of full-texts. We evaluate this question on a large-scale dataset from the medical domain (PubMed) and from economics (EconBiz). In these datasets, the titles and annotations of millions of publications are available, and they outnumber the available full-texts by a factor of 20 and 15, respectively. To exploit these large amounts of data to their full potential, we develop three strong deep learning classifiers and evaluate their performance on the two datasets. The results are promising. On the EconBiz dataset, all three classifiers outperform their full-text counterparts by a large margin. The best title-based classifier outperforms the best full-text method by 9.4%. On the PubMed dataset, the best title-based method almost reaches the performance of the best full-text classifier, with a difference of only 2.9%.

cs.DL↗

arXiv2017

Content Recommendation through Semantic Annotation of User Reviews and Linked Data - An Extended Technical Report

Nowadays, most recommender systems exploit user-provided ratings to infer their preferences. However, the growing popularity of social and e-commerce websites has encouraged users to also share comments and opinions through textual reviews. In this paper, we introduce a new recommendation approach which exploits the semantic annotation of user reviews to extract useful and non-trivial information about the items to recommend. It also relies on the knowledge freely available in the Web of Data, notably in DBpedia and Wikidata, to discover other resources connected with the annotated entities. We evaluated our approach in three domains, using both DBpedia and Wikidata. The results showed that our solution provides a better ranking than another recommendation method based on the Web of Data, while it improves in novelty with respect to traditional techniques based on ratings. Additionally, our method achieved a better performance with Wikidata than DBpedia.

cs.IR↗

arXiv2017

Towards Understanding the Evolution of Vocabulary Terms in Knowledge Graphs

Vocabularies are used for modeling data in Knowledge Graphs (KG) like the Linked Open Data Cloud and Wikidata. During their lifetime, the vocabularies of the KGs are subject to changes. New terms are coined, while existing terms are modified or declared as deprecated. We first quantify the amount and frequency of changes in vocabularies. Subsequently, we investigate to which extend and when the changes are adopted in the evolution of the KGs. We conduct our experiments on three large-scale KGs for which time-stamped snapshots are available, namely the Billion Triples Challenge datasets, Dynamic Linked Data Observatory dataset, and Wikidata. Our results show that the change frequency of terms is rather low, but can have high impact when adopted on a large amount of distributed graph data on the web. Furthermore, not all coined terms are used and most of the deprecated terms are still used by data publishers. There are variations in the adoption time of terms coming from different vocabularies ranging from very fast (few days) to very slow (few years). Surprisingly, there are also adoptions we could observe even before the vocabulary changes are published. Understanding this adoption is important, since otherwise it may lead to wrong assumptions about the modeling status of data published on the web and may result in difficulties when querying the data from distributed sources.

cs.IR↗

arXiv2017

Using Titles vs. Full-text as Source for Automated Semantic Document Annotation

A significant part of the largest Knowledge Graph today, the Linked Open Data cloud, consists of metadata about documents such as publications, news reports, and other media articles. While the widespread access to the document metadata is a tremendous advancement, it is yet not so easy to assign semantic annotations and organize the documents along semantic concepts. Providing semantic annotations like concepts in SKOS thesauri is a classical research topic, but typically it is conducted on the full-text of the documents. For the first time, we offer a systematic comparison of classification approaches to investigate how far semantic annotations can be conducted using just the metadata of the documents such as titles published as labels on the Linked Open Data cloud. We compare the classifications obtained from analyzing the documents' titles with semantic annotations obtained from analyzing the full-text. Apart from the prominent text classification baselines kNN and SVM, we also compare recent techniques of Learning to Rank and neural networks and revisit the traditional methods logistic regression, Rocchio, and Naive Bayes. The results show that across three of our four datasets, the performance of the classifications using only titles reaches over 90% of the quality compared to the classification performance when using the full-text. Thus, conducting document classification by just using the titles is a reasonable approach for automated semantic annotation and opens up new possibilities for enriching Knowledge Graphs.

cs.DL↗

arXiv2016

Profiling vs. Time vs. Content: What does Matter for Top-k Publication Recommendation based on Twitter Profiles? - An Extended Technical Report

So far it is unclear how different factors of a scientific publication recommender system based on users' tweets have an influence on the recommendation performance. We examine three different factors, namely profiling method, temporal decay, and richness of content. Regarding profiling, we compare CF-IDF that replaces terms in TF-IDF by semantic concepts, HCF-IDF as novel hierarchical variant of CF-IDF, and topic modeling. As temporal decay functions, we apply sliding window and exponential decay. In terms of the richness of content, we compare recommendations using both full-texts and titles of publications and using only titles. Overall, the three factors make twelve recommendation strategies. We have conducted an online experiment with 123 participants and compared the strategies in a within-group design. The best recommendations are achieved by the strategy combining CF-IDF, sliding window, and with full-texts. However, the strategies using the novel HCF-IDF profiling method achieve similar results with just using the titles of the publications. Therefore, HCF-IDF can make recommendations when only short and sparse data is available.

cs.DL↗

arXiv2016

TermPicker: Enabling the Reuse of Vocabulary Terms by Exploiting Data from the Linked Open Data Cloud - An Extended Technical Report

Deciding which vocabulary terms to use when modeling data as Linked Open Data (LOD) is far from trivial. Choosing too general vocabulary terms, or terms from vocabularies that are not used by other LOD datasets, is likely to lead to a data representation, which will be harder to understand by humans and to be consumed by Linked data applications. In this technical report, we propose TermPicker: a novel approach for vocabulary reuse by recommending RDF types and properties based on exploiting the information on how other data providers on the LOD cloud use RDF types and properties to describe their data. To this end, we introduce the notion of so-called schema-level patterns (SLPs). They capture how sets of RDF types are connected via sets of properties within some data collection, e.g., within a dataset on the LOD cloud. TermPicker uses such SLPs and generates a ranked list of vocabulary terms for reuse. The lists of recommended terms are ordered by a ranking model which is computed using the machine learning approach Learning To Rank (L2R). TermPicker is evaluated based on the recommendation quality that is measured using the Mean Average Precision (MAP) and the Mean Reciprocal Rank at the first five positions (MRR@5). Our results illustrate an improvement of the recommendation quality by 29% - 36% when using SLPs compared to the beforehand investigated baselines of recommending solely popular vocabulary terms or terms from the same vocabulary. The overall best results are achieved using SLPs in conjunction with the Learning To Rank algorithm Random Forests.

cs.DB↗