Structured Search in Annotated Document Collections

In this work, we describe GYANI (gyan stands for knowledge in Hindi), an indexing infrastructure for search and analysis of large semantically annotated document collections. To facilitate the search for sentences or text regions for many knowledge-centric tasks such as information extraction, question answering, and relationship extraction, it is required that one can query large annotated document collections interactively. However, currently such an indexing infrastructure that scales to millions of documents and provides fast query execution times does not exist. To alleviate this problem, we describe how we can effectively index layers of annotations (e.g., part-of-speech, named entities, temporal expressions, and numerical values) that can be attached to sequences of words. Furthermore, we describe a query language that provides the ability to express regular expressions between word sequences and semantic annotations to ease search for sentences and text regions for enabling knowledge acquisition at scale. We build our infrastructure on a state-of-the-art distributed extensible record store. We extensively evaluate GYANI over two large news archives and the entire Wikipedia amounting to more than fifteen million documents. We observe that using GYANI we can achieve significant speed ups of more than 95x in information extraction, 53x on extracting answer candidates for questions, and 12x on relationship extraction task.

Extracting temporal information from raw text is fundamental for deep language understanding, and key to many applications like question answering, information extraction, and document summarization. Our long-term goal is to build complete temporal structure of documents and apply the temporal structure in other applications like textual entailment, question answering, dialog systems or others. In this paper, we present a first step, a system for extracting event, event features, temporal expression and its normalized values from raw text. Our system is a combination of deep semantic parsing with extraction rules, Markov Logic Network classifiers and Conditional Random Field classifiers. To compare with existing systems, we evaluated our system on the TimeBank corpus. Our system outperforms or does equally well with all existing systems that evaluate on the TimeBank corpus and our performance is very close to inter-annotator agreement of the TimeBank annotators.

Understanding the biology underpinning the natural regeneration of plant species in order to make plans for effective reforestation is a complex task. This can be aided by providing access to databases that contain long-term and wide-scale geographical information on species distribution, habitat, and reproduction. Although there exists widely-used biodiversity databases that contain structured information on species and their occurrences, such as the Global Biodiversity Information Facility (GBIF) and the Atlas of Living Australia (ALA), the bulk of knowledge about biodiversity still remains embedded in textual documents. Unstructured information can be made more accessible and useful for large-scale studies if there are tools and services that automatically extract meaningful information from text and store it in structured formats, e.g., open biodiversity databases, ready to be consumed for analysis (Thessen et al. 2022). We aim to enrich biodiversity occurrence databases with information on species reproductive condition and habitat, derived from text. In previous work, we developed unsupervised approaches to extract related habitats and their locations, and related reproductive condition and temporal expressions (Gabud and Batista-Navarro 2018). We built a new unsupervised hybrid approach for relation extraction (RE), which is a combination of classical rule-based pattern-matching methods and transformer-based language models that framed our RE task as a natural language inference (NLI) task. Using our hybrid approach for RE, we were able to extract related biodiversity entities from text even without a large training dataset. In this work, we implement an information extraction (IE) pipeline comprised of a named entity recognition (NER) tool and our hybrid relation extraction (RE) tool. The NER tool is a transformer-based language model that was pretrained on scientific text and then fine-tuned using COPIOUS (Conserving Philippine Biodiversity by Understanding big data; Nguyen et al. 2019), a gold standard corpus containing named entities relevant to species occurrence. We applied the NER tool to automatically annotate geographical location, temporal expression and habitat information contained within sentences. A dictionary-based approach is then used to identify mentions of reproductive conditions in text (e.g., phrases such as "fruited heavily" and "mass flowering"). We then use our hybrid RE tool to extract reproductive condition - temporal expression and habitat - geographical location entity pairs. We test our IE pipeline on the forestry compendium available in the CABI Digital Library (Centre for Agricultural and Biosciences International), and show that our work enables the enrichment of descriptive information on reproductive and habitat conditions of species. This work is a step towards enhancing a biodiversity database with the inclusion of habitat and reproductive condition information extracted from text.

Accurate analysis of the temporal expression is crucial for Korean text processing applications such as information extraction and chunking for efficient syntactic analysis. It is a complicated problem since temporal expressions often have the ambiguity of syntactic roles. This paper discusses two problems: (1) representing and identifying the temporal expression (2) distinguishing the syntactic function of the temporal expression in case it has a dual syntactic role. In this paper, temporal expressions and the context for disambiguation which is called local context are represented using lexical data extracted from corpus and the finite state transducer. By experiments, it turns out that the method is effective for temporal expression analysis. In particular, our approach shows the corpus-based work could make a promising result for the problem in a restricted domain in that we can effectievely deal with a large size of lexical data.

 Reviews pretty bad fit for the subject matter. ere are not only a great many places to make a start—with Aristotle, with Augustine, with the beginning of the known universe—but these ‘origins’ do not develop in a linear or successive fashion. In many ways there is little difference in the kinds of work done by the essays in the first and second sections of the volume; essays from Part , such as Mark Currie’s delicate, Ricœur-inflected commentary on reading and Tobias Menely’s expansive account of time in ecocriticism, could just as easily have found a place in the ‘Development’ section. ere are other volumes in the Critical Concepts series in which this structure is more natural: Trauma and Literature, edited by J. Roger Kurtz (), is one example that seems to have a more obvious developmental narrative underpinning its narrower critical concept. While I recognize the need for unity across the series and give credit to Allen for bringing creativity to the constraint, the structure does not add much to a collection which could perhaps have developed ideas more usefully between its essays in another order. L H U A B Cyberformalism: Histories of Linguistic Forms in the Digital Archive. By D S. Baltimore: Johns Hopkins University Press. . xxii+ pp. $.. ISBN ––––. e central idea of Daniel Shore’s Cyberformalism is that there exist unified or transferable idea-units called ‘linguistic forms’, and that their discovery and study, using advanced search techniques, can bring us to new understandings of literary history. Linguistic forms can, specifically, help to rewrite the history of textual influence. e ‘linguistic form’, explains Shore, is a type of sign-unit: it is an unfilled or partially filled construction. A linguistic form is not empty syntax, nor is it a Saussurean sign, but rather it exists somewhere in between. ese forms persist through time and space but, as with words, they change, adapt, and take on new meaning according to local context. ‘What Would Jesus Do?’ is an example of a filled-in form: Shore explains that while the specific phrase originated in the nineteenth century, it can be traced back to the earliest Christian theology, and he argues that one can detect, in the eighteenth and nineteenth centuries, a shi from ‘Imitatio Christi’ to the conditional phrase which eventually became the form we know and use today. Digital archives offer opportunities for the study of the linguistic form, but finding them requires smart tools and researchers. Searching for words or sequences of words does not help to find instances of a linguistic form, because by definition they are lexicographically distinct. To combat this Shore uses advanced search techniques , which he outlines throughout the text. Two underlying technologies allow him to produce results from complex queries: ‘Part of Speech’ tagging and Regular Expressions (computer code developed specifically for matching complicated patterns in text). Using these in tandem enables the detection of forms even through extensive syntactic variation. is is exciting work: Shore’s process of discovery reads, in part, as detective fiction. e forms, we discover, take on multiple roles MLR, .,   and change in meaning depending on context: in further case studies Shore traces a network of poetic influence through the phrase ‘Was it for this’, and shows how the form version of ‘act as if [x]’ was used by early advocates in the establishment of a nineteenth-century ‘technology of the self’, and also by Kant to ‘preserve a practical role for the metaphysical ideas that his epistemology places beyond the bounds of knowledge’ (p. ). Perhaps the most important contribution made by this book is to the study of style: Shore’s fascinating work on Milton’s ‘depictives’ (adjectives which form part of the predicate but modify the subject, as in ‘Death | Grinned horrible’ (Paradise Lost, . –)) is followed by a wonderful examination of the ways in which advanced search can move forward our understanding of style, iterability, and the evolution or repurposing of work or ideas. rough the ambiguity inherent in a depictive form we gain a greater understanding of Milton’s work: through detective work Shore discovered from whom Milton appropriated...

The field of automatic speech recognition (ASR) has enjoyed more than 30 years of technology advances due to the extensive utilization of the hidden Markov model (HMM) framework and a concentrated effort by the speech community to make available a vast amount of speech and language resources, known today as the Big Data Paradigm. State-of-the-art ASR systems achieve a high recognition accuracy for well-formed utterances of a variety of languages by decoding speech into the most likely sequence of words among all possible sentences represented by a finite-state network (FSN) approximation of all the knowledge sources required by the ASR task. However, the ASR problem is still far from being solved because not all information available in the speech knowledge hierarchy can be directly integrated into the FSN to improve the ASR performance and enhance system robustness. It is believed that some of the current issues of integrating various knowledge sources in top-down integrated search can be partially addressed by processing techniques that take advantage of the full set of acoustic and language information in speech. It has long been postulated that human speech recognition (HSR) determines the linguistic identity of a sound based on detected evidence that exists at various levels of the speech knowledge hierarchy, ranging from acoustic phonetics to syntax and semantics. This calls for a bottom-up attribute detection and knowledge integration framework that links speech processing with information extraction, by spotting speech cues with a bank of attribute detectors, weighting and combining acoustic evidence to form cognitive hypotheses, and verifying these theories until a consistent recognition decision can be reached. The recently proposed automatic speech attribute transcription (ASAT) framework is an attempt to mimic some HSR capabilities with asynchronous speech event detection followed by bottom-up knowledge integration and verification. In the last few years, ASAT has demonstrated good potential and has been applied to a variety of existing applications in speech processing and information extraction.

Neural network language models, such as BERT, can be used for information extraction from medical texts with unstructured free text. These models can be pre-trained on a large corpus to learn the language and characteristics of the relevant domain and then fine-tuned with labeled data for a specific task. We propose a pipeline using human-in-the-loop labeling to create annotated data for Estonian healthcare information extraction. This method is particularly useful for low-resource languages and is more accessible to those in the medical field than rule-based methods like regular expressions.

Recognizing and extracting meaningful information from Web unstructured documents, taking into account their semantics, is an important problem of information and knowledge management. This paper describes H\(\imath\)LeX, a system implementing a novel logic-based approach to information extraction from unstructured documents. The approach adopted in the H\(\imath\)LeX system is founded on a new two-dimensional representation of documents, and heavily exploits DLP + - an extension of disjunctive logic programming for ontology representation and reasoning, which has been recently implemented on top of the DLV reasoning environment. Unlike previous systems, which are mainly syntactic, H\(\imath\)LeX combines both semantic and syntactic knowledge for a powerful information extraction. Ontologies, representing the semantics of information to be extracted, are encoded in DLP + , while the extraction patterns are expressed using regular expressions and an ad hoc two-dimensional grammar. The execution of DLP + reasoning modules, encoding the grammar expressions, yields the actual extraction of information from the input document. H\(\imath\)LeX allows the semantic information extraction from both HTML pages and flat text documents by using synthetic and very expressive extraction patterns.

In today's environment of information overload, Question Answering (QA) is a critically important research area. QA is the task of automatically extracting a precise answer from one or more data sources to a question posed in natural language. A twostage strategy is typically adopted when designing a QA system; the first stage is an Information Retrieval (IR) process which returns a set of candidate documents relevant to the question and the second stage narrows the information contained in those passages down to a single response (sentence or entity) that answers the question, typically using Information Extraction (IE) or Natural Language Processing methods. This research proposes novel techniques for QA by enhancing the user's original query with latent semantic information from the corpus. This enhanced query is then applied to both the first and second stages of the QA architecture. To build the enhanced query, we propose the Aspect-Based Relevance Language Model as an approach that uses statistical language modeling techniques to measure the likelihood of relevance of a concept (oraspect as defined by Probabilistic Latent Semantic Analysis) to a question. We then use terms from the aspects that have the highest likelihood of relevance to design a model for a semantic Question Context, which includes sense-disambiguated terms than amplify the user's query. Question Context is incorporated into the first state of QA as query expansion to improve recall. We then derive a novel measure called Answer Credibility from the Question Context. Answer Credibility may be thought of as a statistical measure of the reliability of a candidate answer with respect to a question and the source text from which the candidate answer was derived. We incorporate Answer Credibility in the Answer Validation process; the answer with the highest score after the application of Answer Credibility is returned to the user. Our techniques show performance improvements over state-of-the-art approaches, and have the advantage that they use statistical techniques to derive semantic information to aid the process of QA.

Memory-Based Learning (MBL), one of the most influential machine learning paradigms, has been applied with great success to a variety of NLP tasks. This monograph describes the application of MBL to robust parsing. Robust parsing using MBL can provide added functionality for key NLP applications, such as Information Retrieval, Information Extraction, and Question Answering, by facilitating more complex syntactic analysis than is currently available. The text presupposes no prior knowledge of MBL. It provides a comprehensive introduction to the framework and goes on to describe and compare applications of MBL to parsing. Since parsing is not easily characterizable as a classification task, adaptations of standard MBL are necessary. These adaptations can either take the form of a cascade of local classifiers or of a holistic approach for selecting a complete tree. The text provides excellent course material on MBL. It is equally relevant for any researcher concerned with symbolic machine learning, Information Retrieval, Information Extraction, and Question Answering.

Information extraction and question answering have the potential to introduce a new paradigm for how machine learning is applied to criminal law. Existing approaches generally use tabular data for predictive metrics. An alternative approach is needed for matters of equitable justice, where individuals are judged on a case-by-case basis, in a process involving verbal or written discussion and interpretation of case factors. Such discussions are individualized, but they nonetheless rely on underlying facts. Information extraction can play an important role in surfacing these facts, which are still important to understand. We analyze unsupervised, weakly supervised, and pre-trained models’ ability to extract such factual information from the free-form dialogue of California parole hearings. With a few exceptions, most F1 scores are below 0.85. We use this opportunity to highlight some opportunities for further research for information extraction and question answering. We encourage new developments in NLP to enable analysis and review of legal cases to be done in a post-hoc, not predictive, manner.

Machine-readable semantic knowledge in the form of taxonomies (i.e., a collection of is-a edges) has proved to be beneficial in an array of NLP tasks including inference, textual entailment, question answering and information extraction. Such widespread utility of taxonomies has led to multiple large-scale manual efforts towards taxonomy induction such as WordNet and Cyc. However, manual construction of taxonomies is time-intensive, and usually, requires substantial annotation efforts by domain experts. Furthermore, the resulting taxonomies suffer from low coverage and are unavailable for specific domains or languages. Therefore, in recent years, there has been a growing body of work, which aims to induce taxonomies automatically, either from semi-structured knowledge resources or unstructured text. In this thesis, we focus on the task of automated taxonomy induction under a variety of different settings. We first focus on taxonomy induction from the largest semi-structured knowledge resource, i.e., Wikipedia. More specifically, we introduce a set of novel heuristics aimed towards inducing a large-scale taxonomy from the English Wikipedia categories network. We also propose a novel comprehensive path-based evaluation framework for taxonomies. Taxonomy induced using our approach significantly outperforms the state of the art across both edge-based as well as path-based evaluation metrics. Moreover, our experiments also demonstrate that high accuracy in edge-based evaluation metrics does not always translate to high-accuracy in path-based evaluation metrics. Subsequently, we propose a novel approach, which leverages the interlanguage links of Wikipedia to induce taxonomies in other Wikipedia languages. Compared to the state of the art, our approach is simpler, more principled, and results in taxonomies that are significantly more accurate across both edge-based and path-based evaluation metrics. In the second part of this thesis, we focus on taxonomy induction from unstructured text. We propose a novel approach towards taxonomy induction from an input vocabulary of seed terms. Unlike previous approaches, which extract singular hypernym edges for terms, our approach utilizes a novel probabilistic framework to extract long-range hypernym subsequences. Taxonomy induction from the extracted subsequences is cast as an instance of the minimum-cost flow optimization problem on a carefully-designed flow network. Through experiments, we demonstrate that our approach outperforms the state of the art across four languages. We also show that our approach is robust to the presence of noise in the input vocabulary. Our approach facilitates the relaxation of many simplifying assumptions, which were employed by previous taxonomy induction approaches. As a result, our work serves to automate the process of taxonomy induction from unstructured text in the true sense. Finally, we introduce a task of discovering and generalizing lexicalized templates from the titles of Wikipedia entities. The experimental results on this task demonstrate that taxonomies, which perform better on our proposed path-based metrics, result in a more accurate set of generalizations for a given set of entities. In summary, this thesis proposes new approaches towards automated taxonomy induction. It improves upon the state of the art in a variety of different settings. It also relaxes many simplifying assumptions that limited the applicability of prior approaches.

This paper describes the KnowledgeHub tool, a scientific literature Information Extraction (IE) and Question Answering (QA) pipeline. This is achieved by supporting the ingestion of PDF documents that are converted to text and structured representations. An ontology can then be constructed where a user defines the types of entities and relationships they want to capture. A browser-based annotation tool enables annotating the contents of the PDF documents according to the ontology. Named Entity Recognition (NER) and Relation Classification (RC) models can be trained on the resulting annotations and can be used to annotate the unannotated portion of the documents. A knowledge graph is constructed from these entity and relation triples which can be queried to obtain insights from the data. Furthermore, we integrate a suite of Large Language Models (LLMs) that can be used for QA and summarisation that is grounded in the included documents. KnowledgeHub is a unique tool that supports annotation, IE and QA, which gives the user full insight into the knowledge discovery pipeline.

This paper proposes an information extraction model that identifies text patterns representing relations between two entities. It is proposed that, given a set of entity pairs representing a specific relation, it is possible to find text patterns representing such relation within sentences from documents containing those entites. After those text patterns are identified, it is possible to attempt the extraction of a complementary entity, considering the first entity of the relation and the related text patterns are provided. The pattern selection relies on regular expressions, frequency and identification of less relevant words. Modern search engines APIs and HTML parsers are used to retrieve and parse web pages in real time, eliminating the need of a pre-established corpus. The retrieval of document counts within a timeframe is also used to aid in the selection of the entities extracted.

Information extraction is concerned with applying natural language processing to automatically extract the essential details from text documents. A great disadvantage of current approaches is their intrinsic dependence to the application domain and the target language. Several machine learning techniques have been applied in order to facilitate the portability of the information extraction systems. This paper describes a general method for building an information extraction system using regular expressions along with supervised learning algorithms. In this method, the extraction decisions are lead by a set of classifiers instead of sophisticated linguistic analyses. The paper also shows a system called TOPO that allows to extract the information related with natural disasters from newspaper articles in Spanish language. Experimental results of this system indicate that the proposed method can be a practical solution for building information extraction systems reaching an F-measure as high as 72%.KeywordsNatural Language ProcessingRegular ExpressionInformation ExtractionNews ReportMachine Learn ApproachThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.