Probabilistic information retrieval model for a dependency structured indexing system

Most previous information retrieval (IR) models assume that terms of queries and documents are statistically independent from each other. However, conditional independence assumption is obviously and openly understood to be wrong, so we present a new method of incorporating term dependence into a probabilistic retrieval model by adapting a dependency structured indexing system using a dependency parse tree and Chow Expansion to compensate the weakness of the assumption. In this paper, we describe a theoretic process to apply the Chow Expansion to the general probabilistic models and the state-of-the-art 2-Poisson model. Through experiments on document collections in English and Korean, we demonstrate that the incorporation of term dependences using Chow Expansion contributes to the improvement of performance in probabilistic IR systems.     

An efficient document clustering algorithm and its application to a document browser

We present an efficient document clustering algorithm that uses a term frequency vector for each document instead of using a huge proximity matrix. The algorithm has the following features: (1) it requires a relatively small amount of memory and runs fast, (2) it produces a hierarchy in the form of a document classification tree and (3) the hierarchy obtained by the algorithm explicitly reveals a collection structure. We confirm these features and thus show the algorithm's feasibility through clustering experiments in which we use two collections of Japanese documents, the sizes of which are 83,099 and 14,701 documents. We also introduce an application of this algorithm to a document browser. This browser is used in our Japanese-to-English translation aid system. The browsing module of the system consists of a huge database of Japanese news articles and their English translations. The Japanese article collection is clustered into a hierarchy by our method. Since each node in the hierarchy corresponds to a topic in the collection, we can use the hierarchy to directly access articles by topic. A user can learn general translation knowledge of each topic by browsing the Japanese articles and their English translations. We also discuss techniques of presenting a large tree-formed hierarchy on a computer screen.     

An exploration of ranking models and feedback method for related entity finding

Most existing search engines focus on document retrieval. However, information needs are certainly not limited to finding relevant documents. Instead, a user may want to find relevant entities such as persons and organizations. In this paper, we study the problem of related entity finding. Our goal is to rank entities based on their relevance to a structured query, which specifies an input entity, the type of related entities and the relation between the input and related entities. We first discuss a general probabilistic framework, derive six possible retrieval models to rank the related entities, and then compare these models both analytically and empirically. To further improve performance, we study the problem of feedback in the context of related entity finding. Specifically, we propose a mixture model based feedback method that can utilize the pseudo feedback entities to estimate an enriched model for the relation between the input and related entities. Experimental results over two standard TREC collections show that the derived relation generation model combined with a relation feedback method performs better than other models.     

Introducing structure management in automatic reference resolution: An XML-based approach

References to parts of structured documents use their structure to locate the piece of document which is the reference target. On the other hand, XML has become an increasingly important language for structured documents. One of its most important related languages is XPath, the language that permits fragments of XML documents to be selected. In this article we present a methodology, and an application case, to automatically extract and solve references to fragments of structured documents. This approach combines structure manipulation and information extraction, to enhance reference extraction tools by improving the precision of the references extracted. We take advantage of XML markup to locate the position within the structure in which the references are found. The use of XPath, one of the most important XML related languages, for reference resolution is original: the resolution tool automatically builds XPath expressions. This proposal is inspired (and implemented) from our work with legislative documents.     

Learning entity-centric document representations using an entity facet topic model

Learning semantic representations of documents is essential for various downstream applications, including text classification and information retrieval. Entities, as important sources of information, have been playing a crucial role in assisting latent representations of documents. In this work, we hypothesize that entities are not monolithic concepts; instead they have multiple aspects, and different documents may be discussing different aspects of a given entity. Given that, we argue that from an entity-centric point of view, a document related to multiple entities shall be (a) represented differently for different entities (multiple entity-centric representations), and (b) each entity-centric representation should reflect the specific aspects of the entity discussed in the document.In this work, we devise the following research questions: (1) Can we confirm that entities have multiple aspects, with different aspects reflected in different documents, (2) can we learn a representation of entity aspects from a collection of documents, and a representation of document based on the multiple entities and their aspects as reflected in the documents, (3) does this novel representation improves algorithm performance in downstream applications, and (4) what is a reasonable number of aspects per entity? To answer these questions we model each entity using multiple aspects (entity facets¹), where each entity facet is represented as a mixture of latent topics. Then, given a document associated with multiple entities, we assume multiple entity-centric representations, where each entity-centric representation is a mixture of entity facets for each entity. Finally, a novel graphical model, the Entity Facet Topic Model (EFTM), is proposed in order to learn entity-centric document representations, entity facets, and latent topics.Through experimentation we confirm that (1) entities are multi-faceted concepts which we can model and learn, (2) a multi-faceted entity-centric modeling of documents can lead to effective representations, which (3) can have an impact in downstream application, and (4) considering a small number of facets is effective enough. In particular, we visualize entity facets within a set of documents, and demonstrate that indeed different sets of documents reflect different facets of entities. Further, we demonstrate that the proposed entity facet topic model generates better document representations in terms of perplexity, compared to state-of-the-art document representation methods. Moreover, we show that the proposed model outperforms baseline methods in the application of multi-label classification. Finally, we study the impact of EFTM’s parameters and find that a small number of facets better captures entity specific topics, which confirms the intuition that on average an entity has a small number of facets reflected in documents.     

Mining fuzzy frequent itemsets for hierarchical document clustering

As text documents are explosively increasing in the Internet, the process of hierarchical document clustering has been proven to be useful for grouping similar documents for versatile applications. However, most document clustering methods still suffer from challenges in dealing with the problems of high dimensionality, scalability, accuracy, and meaningful cluster labels. In this paper, we will present an effective Fuzzy Frequent Itemset-Based Hierarchical Clustering (F²IHC) approach, which uses fuzzy association rule mining algorithm to improve the clustering accuracy of Frequent Itemset-Based Hierarchical Clustering (FIHC) method. In our approach, the key terms will be extracted from the document set, and each document is pre-processed into the designated representation for the following mining process. Then, a fuzzy association rule mining algorithm for text is employed to discover a set of highly-related fuzzy frequent itemsets, which contain key terms to be regarded as the labels of the candidate clusters. Finally, these documents will be clustered into a hierarchical cluster tree by referring to these candidate clusters. We have conducted experiments to evaluate the performance based on Classic4, Hitech, Re0, Reuters, and Wap datasets. The experimental results show that our approach not only absolutely retains the merits of FIHC, but also improves the accuracy quality of FIHC.     

Using language models to improve opinion detection

Opinion mining is one of the most important research tasks in the information retrieval research community. With the huge volume of opinionated data available on the Web, approaches must be developed to differentiate opinion from fact. In this paper, we present a lexicon-based approach for opinion retrieval. Generally, opinion retrieval consists of two stages: relevance to the query and opinion detection. In our work, we focus on the second state which itself focusses on detecting opinionated documents . We compare the document to be analyzed with opinionated sources that contain subjective information. We hypothesize that a document with a strong similarity to opinionated sources is more likely to be opinionated itself. Typical lexicon-based approaches treat and choose their opinion sources according to their test collection, then calculate the opinion score based on the frequency of subjective terms in the document. In our work, we use different open opinion collections without any specific treatment and consider them as a reference collection. We then use language models to determine opinion scores. The analysis document and reference collection are represented by different language models (i.e., Dirichlet, Jelinek-Mercer and two-stage models). These language models are generally used in information retrieval to represent the relationship between documents and queries. However, in our study, we modify these language models to represent opinionated documents. We carry out several experiments using Text REtrieval Conference (TREC) Blogs 06 as our analysis collection and Internet Movie Data Bases (IMDB), Multi-Perspective Question Answering (MPQA) and CHESLY as our reference collection. To improve opinion detection, we study the impact of using different language models to represent the document and reference collection alongside different combinations of opinion and retrieval scores. We then use this data to deduce the best opinion detection models. Using the best models, our approach improves on the best baseline of TREC Blog (baseline4) by 30%.     

A novel reasoning mechanism for multi-label text classification

The aim in multi-label text classification is to assign a set of labels to a given document. Previous classifier-chain and sequence-to-sequence models have been shown to have a powerful ability to capture label correlations. However, they rely heavily on the label order, while labels in multi-label data are essentially an unordered set. The performance of these approaches is therefore highly variable depending on the order in which the labels are arranged. To avoid being dependent on label order, we design a reasoning-based algorithm named Multi-Label Reasoner (ML-Reasoner) for multi-label classification. ML-Reasoner employs a binary classifier to predict all labels simultaneously and applies a novel iterative reasoning mechanism to effectively utilize the inter-label information, where each instance of reasoning takes the previously predicted likelihoods for all labels as additional input. This approach is able to utilize information between labels, while avoiding the issue of label-order sensitivity. Extensive experiments demonstrate that our method outperforms state-of-the art approaches on the challenging AAPD dataset. We also apply our reasoning module to a variety of strong neural-based base models and show that it is able to boost performance significantly in each case.     

Managing structured queries in probabilistic XML retrieval systems

Focusing on the context of XML retrieval, in this paper we propose a general methodology for managing structured queries (involving both content and structure) within any given structured probabilistic information retrieval system which is able to compute posterior probabilities of relevance for structural components given a non-structured query (involving only query terms but not structural restrictions). We have tested our proposal using two specific information retrieval systems (Garnata and PF/Tijah), and the structured document collections from the last six editions of the INitiative for the Evaluation of XML Retrieval (INEX).     

The challenge of commercial document retrieval,Part I: Major issues,and a framework based on search exhaustivity,determinacy of representation and document collection size

With the growing focus on what is collectively known as “knowledge management”, a shift continues to take place in commercial information system development: a shift away from the well-understood data retrieval/database model, to the more complex and challenging development of commercial document/information retrieval models. While document retrieval has had a long and rich legacy of research, its impact on commercial applications has been modest. At the enterprise level most large organizations have little understanding of, or commitment to, high quality document access and management. Part of the reason for this is that we still do not have a good framework for understanding the major factors which affect the performance of large-scale corporate document retrieval systems. The thesis of this discussion is that document retrieval—specifically, access to intellectual content—is a complex process which is most strongly influenced by three factors: the size of the document collection; the type of search (exhaustive, existence or sample); and, the determinacy of document representation. Collectively, these factors can be used to provide a useful framework for, or taxonomy of, document retrieval, and highlight some of the fundamental issues facing the design and development of commercial document retrieval systems. This is the first of a series of three articles. Part II (D.C. Blair, The challenge of commercial document retrieval. Part II. A strategy for document searching based on identifiable document partitions, Information Processing and Management, 2001b, this issue) will discuss the implications of this framework for search strategy, and Part III (D.C. Blair, Some thoughts on the reported results of Text REtrieval Conference (TREC), Information Processing and Management, 2002, forthcoming) will consider the importance of the TREC results for our understanding of operating information retrieval systems.     

Multiple sets of features for automatic genre classification of web documents

With the increase of information on the Web, it is difficult to find desired information quickly out of the documents retrieved by a search engine. One way to solve this problem is to classify web documents according to various criteria. Most document classification has been focused on a subject or a topic of a document. A genre or a style is another view of a document different from a subject or a topic. The genre is also a criterion to classify documents. In this paper, we suggest multiple sets of features to classify genres of web documents. The basic set of features, which have been proposed in the previous studies, is acquired from the textual properties of documents, such as the number of sentences, the number of a certain word, etc. However, web documents are different from textual documents in that they contain URL and HTML tags within the pages. We introduce new sets of features specific to web documents, which are extracted from URL and HTML tags. The present work is an attempt to evaluate the performance of the proposed sets of features, and to discuss their characteristics. Finally, we conclude which is an appropriate set of features in automatic genre classification of web documents.     

Probabilistic topic modeling in multilingual settings: An overview of its methodology and applications

Probabilistic topic models are unsupervised generative models which model document content as a two-step generation process, that is, documents are observed as mixtures of latent concepts or topics, while topics are probability distributions over vocabulary words. Recently, a significant research effort has been invested into transferring the probabilistic topic modeling concept from monolingual to multilingual settings. Novel topic models have been designed to work with parallel and comparable texts. We define multilingual probabilistic topic modeling (MuPTM) and present the first full overview of the current research, methodology, advantages and limitations in MuPTM. As a representative example, we choose a natural extension of the omnipresent LDA model to multilingual settings called bilingual LDA (BiLDA). We provide a thorough overview of this representative multilingual model from its high-level modeling assumptions down to its mathematical foundations. We demonstrate how to use the data representation by means of output sets of (i) per-topic word distributions and (ii) per-document topic distributions coming from a multilingual probabilistic topic model in various real-life cross-lingual tasks involving different languages, without any external language pair dependent translation resource: (1) cross-lingual event-centered news clustering, (2) cross-lingual document classification, (3) cross-lingual semantic similarity, and (4) cross-lingual information retrieval. We also briefly review several other applications present in the relevant literature, and introduce and illustrate two related modeling concepts: topic smoothing and topic pruning. In summary, this article encompasses the current research in multilingual probabilistic topic modeling. By presenting a series of potential applications, we reveal the importance of the language-independent and language pair independent data representations by means of MuPTM. We provide clear directions for future research in the field by providing a systematic overview of how to link and transfer aspect knowledge across corpora written in different languages via the shared space of latent cross-lingual topics, that is, how to effectively employ learned per-topic word distributions and per-document topic distributions of any multilingual probabilistic topic model in various cross-lingual applications.     

Abstractive headline generation using WIDL-expressions

We present a new paradigm for the automatic creation of document headlines that is based on direct transformation of relevant textual information into well-formed textual output. Starting from an input document, we automatically create compact representations of weighted finite sets of strings, called WIDL-expressions, which encode the most important topics in the document. A generic natural language generation engine performs the headline generation task, driven by both statistical knowledge encapsulated in WIDL-expressions (representing topic biases induced by the input document) and statistical knowledge encapsulated in language models (representing biases induced by the target language). Our evaluation shows similar performance in quality with a state-of-the-art, extractive approach to headline generation, and significant improvements in quality over previously proposed solutions to abstractive headline generation.     

Web document summarization by exploiting social context with matrix co-factorization

In the context of social media, users usually post relevant information corresponding to the contents of events mentioned in a Web document. This information posses two important values in that (i) it reflects the content of an event and (ii) it shares hidden topics with sentences in the main document. In this paper, we present a novel model to capture the nature of relationships between document sentences and post information (comments or tweets) in sharing hidden topics for summarization of Web documents by utilizing relevant post information. Unlike previous methods which are usually based on hand-crafted features, our approach ranks document sentences and user posts based on their importance to the topics. The sentence-user-post relation is formulated in a share topic matrix, which presents their mutual reinforcement support. Our proposed matrix co-factorization algorithm computes the score of each document sentence and user post and extracts the top ranked document sentences and comments (or tweets) as a summary. We apply the model to the task of summarization on three datasets in two languages, English and Vietnamese, of social context summarization and also on DUC 2004 (a standard corpus of the traditional summarization task). According to the experimental results, our model significantly outperforms the basic matrix factorization and achieves competitive ROUGE-scores with state-of-the-art methods.     

Soft-constrained inference for Named Entity Recognition

Much of the valuable information in supporting decision making processes originates in text-based documents. Although these documents can be effectively searched and ranked by modern search engines, actionable knowledge need to be extracted and transformed in a structured form before being used in a decision process. In this paper we describe how the discovery of semantic information embedded in natural language documents can be viewed as an optimization problem aimed at assigning a sequence of labels (hidden states) to a set of interdependent variables (textual tokens). Dependencies among variables are efficiently modeled through Conditional Random Fields, an indirected graphical model able to represent the distribution of labels given a set of observations. The Markov property of these models prevent them to take into account long-range dependencies among variables, which are indeed relevant in Natural Language Processing. In order to overcome this limitation we propose an inference method based on Integer Programming formulation of the problem, where long distance dependencies are included through non-deterministic soft constraints.     

Applying regression models to query-focused multi-document summarization

Most existing research on applying machine learning techniques to document summarization explores either classification models or learning-to-rank models. This paper presents our recent study on how to apply a different kind of learning models, namely regression models, to query-focused multi-document summarization. We choose to use Support Vector Regression (SVR) to estimate the importance of a sentence in a document set to be summarized through a set of pre-defined features. In order to learn the regression models, we propose several methods to construct the “pseudo” training data by assigning each sentence with a “nearly true” importance score calculated with the human summaries that have been provided for the corresponding document set. A series of evaluations on the DUC data sets are conducted to examine the efficiency and the robustness of the proposed approaches. When compared with classification models and ranking models, regression models are consistently preferable.     

Document indexing: a concept-based approach to term weight estimation

Traditional index weighting approaches for information retrieval from texts depend on the term frequency based analysis of the text contents. A shortcoming of these indexing schemes, which consider only the occurrences of the terms in a document, is that they have some limitations in extracting semantically exact indexes that represent the semantic content of a document. To address this issue, we developed a new indexing formalism that considers not only the terms in a document, but also the concepts. In this approach, concept clusters are defined and a concept vector space model is proposed to represent the semantic importance degrees of lexical items and concepts within a document. Through an experiment on the TREC collection of Wall Street Journal documents, we show that the proposed method outperforms an indexing method based on term frequency (TF), especially in regard to the few highest-ranked documents. Moreover, the index term dimension was 80% lower for the proposed method than for the TF-based method, which is expected to significantly reduce the document search time in a real environment.     

Centroid-based summarization of multiple documents

We present a multi-document summarizer, MEAD, which generates summaries using cluster centroids produced by a topic detection and tracking system. We describe two new techniques, a centroid-based summarizer, and an evaluation scheme based on sentence utility and subsumption. We have applied this evaluation to both single and multiple document summaries. Finally, we describe two user studies that test our models of multi-document summarization.     

Formal language models for finding groups of experts

The task of finding groups or teams has recently received increased attention, as a natural and challenging extension of search tasks aimed at retrieving individual entities. We introduce a new group finding task: given a query topic, we try to find knowledgeable groups that have expertise on that topic. We present five general strategies for this group finding task, given a heterogenous document repository. The models are formalized using generative language models. Two of the models aggregate expertise scores of the experts in the same group for the task, one locates documents associated with experts in the group and then determines how closely the documents are associated with the topic, whilst the remaining two models directly estimate the degree to which a group is a knowledgeable group for a given topic. For evaluation purposes we construct a test collection based on the TREC 2005 and 2006 Enterprise collections, and define three types of ground truth for our task. Experimental results show that our five knowledgeable group finding models achieve high absolute scores. We also find significant differences between different ways of estimating the association between a topic and a group.     

Global information-aware argument mining based on a top-down multi-turn QA model

Argument mining (AM) aims to automatically generate a graph that represents the argument structure of a document. Most previous AM models only pay attention to a single argument component (AC) to classify the type of the AC or a pair of ACs to identify and classify the argumentative relation (AR) between the two ACs. These models ignore the impact of global argument structure of the documents, which is important, especially in some highly structured genres such as scientific papers, where the process of argumentation is relatively fixed. Inspired by this, we propose a novel two-stage model which leverages global structure information to support AM. The first stage uses a multi-turn question-answering model to incrementally generate an initial argumentative graph that identifies relations among ACs. At each turn, all ACs related to the query AC are generated simultaneously, such that the sibling global information between the answer ACs is considered. In addition, the partially constructed graph is used as global structure information to support the extension of the graph with additional ACs. After the whole initial graph structure has been determined, the second stage assigns semantic types to both the ACs and ARs among them, leveraging information from this initial graph as global structure information. We test the proposed methods on two scientific datasets (one is the AbstRCT dataset including 659 abstracts about cancer research and the other is the SciARG dataset that consists of 225 computer linguistic abstracts and 285 biomedical abstracts) and a student essay dataset PE with 402 essays. Our experiments show that our model improves the state-of-the-art performance on two scientific datasets for different AM subtasks, with average improvements of 1%, 2.41%, 1.1% for the ACC, ARI and ARC task respectively on the AbstRCT dataset, and 2.36%, 1.84%, 8.87% for the ACC, ARI and ARC task on the SciARG dataset. Our model also achieves comparative results on the PE datasets: 87.7% of F1 scores for the ACC task, 81.4% for the ARI task and 78.8% for the ARC task.