Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 94 views

    Text Mining Research Papers PDF

    Uploaded by

    Javer Rosqueta Cañaveral
    Text mining is the process of extracting meaningful information from natural language text. It involves analyzing text to extract useful information for particular purposes. Research papers contain the most complete representation of human knowledge. Recent developments in machine learning, computing power and data storage, as well as increased access to research papers through open access initiatives, have reduced barriers to text mining of research papers. Text mining of research papers can support applications such as literature based discovery, summarization of research findings, question answering, and understanding research trends. A typical text mining workflow involves extracting and cleaning text, followed by tokenization, stemming, lemmatization, parsing, and named entity recognition to derive semantic information that can be analyzed and visualized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Text Mining Research Papers PDF

    Uploaded by

    Javer Rosqueta Cañaveral
    94 views28 pages
    Text mining is the process of extracting meaningful information from natural language text. It involves analyzing text to extract useful information for particular purposes. Research papers contain the most complete representation of human knowledge. Recent developments in machine learning, computing power and data storage, as well as increased access to research papers through open access initiatives, have reduced barriers to text mining of research papers. Text mining of research papers can support applications such as literature based discovery, summarization of research findings, question answering, and understanding research trends. A typical text mining workflow involves extracting and cleaning text, followed by tokenization, stemming, lemmatization, parsing, and named entity recognition to derive semantic information that can be analyzed and visualized.

    Original Title

    Text-Mining-Research-Papers.pptx.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Text mining is the process of extracting meaningful information from natural language text. It involves analyzing text to extract useful information for particular purposes. Research papers contain the most complete representation of human knowledge. Recent developments in machine learning, computing power and data storage, as well as increased access to research papers through open access initiatives, have reduced barriers to text mining of research papers. Text mining of research papers can support applications such as literature based discovery, summarization of research findings, question answering, and understanding research trends. A typical text mining workflow involves extracting and cleaning text, followed by tokenization, stemming, lemmatization, parsing, and named entity recognition to derive semantic information that can be analyzed and visualized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    94 views28 pages

    Text Mining Research Papers PDF

    Uploaded by

    Javer Rosqueta Cañaveral
    Text mining is the process of extracting meaningful information from natural language text. It involves analyzing text to extract useful information for particular purposes. Research papers contain the most complete representation of human knowledge. Recent developments in machine learning, computing power and data storage, as well as increased access to research papers through open access initiatives, have reduced barriers to text mining of research papers. Text mining of research papers can support applications such as literature based discovery, summarization of research findings, question answering, and understanding research trends. A typical text mining workflow involves extracting and cleaning text, followed by tokenization, stemming, lemmatization, parsing, and named entity recognition to derive semantic information that can be analyzed and visualized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 28

    An Introduction to Text Mining

    Research Papers
    Petr Knoth
    Phil Gooch
    Mendeley
    22 September 2015
    What is text mining?
    ‘the process of deriving high-quality information from text’ (Wikipedia)

    ‘the discovery by computer of new, previously unknown information, by


    automatically extracting information from different written resources. A key
    element is the linking ... of the extracted information ... to form new facts or
    new hypotheses to be explored further’ (Hearst, 2003)

    ‘a burgeoning new field that attempts to glean meaningful information from


    natural language text ... the process of analyzing text to extract information
    that is useful for particular purposes’
    (Witten, 2004)
    Why text mine research papers?

    “Research papers are the most complete representation of human knowledge.“


    Why is it so much talked about now?
    The idea is not new, but up until recently access to large amounts of research
    papers was controlled by a handful of companies having bespoke
    arrangements with publishers.

    The Open Access movement has recently largely contributed to decreasing


    the barriers to text-mining of research papers.

    The availability of tools, developments in machine learning, and reduction in


    the costs of computing power and storage, has removed some of the technical
    and financial barriers.
    What are the opportunities of text
    mining research papers?
    • Literature Based Discovery (LBD)
    • Undiscovered Public Knowledge (Swanson, 1986).
    • Mining relationships for which there is “hidden” evidence
    in the research literature, yet they are not explicitly
    stated, such as magnesium deficiency and migraine, fish
    oil and Raynaud's disease.
    • Swanson's discoveries simulated by automated
    techniques (Weeber et al., 2001).
    Other use cases
    • Supporting exploratory access to research literature
    • staying up-to-date with research
    • analysing, comparing and contrasting research findings
    • Summarisation of research findings
    • Systematic literature review automation
    • ‘snowballing’
    • Question answering and semantic search from papers
    • Understanding the research impact of articles, individuals,
    institutions, countries, …
    • Monitoring research trends
    • Understanding how to direct research funding …
    • Evidence of reuse and plagiarism detection …
    Generic text mining workflow
    Extraction and Sentence
    Tokenisation
    Papers text cleaning splitting

    Stemming, Part of Speech


    lemmatisation Parsing
    Tagging (POS)

    Named Entity
    Relation extraction
    Recognition (NER)

    Semantic Analysis
    search, Knowledge and
    reasoning, ... base visualisation
    Example 1: Literature Based Discovery
    ● A range of techniques (Smalheiser, 2012)
    ● A typical approach: ABC method (e.g. Hristovski et. al. 2008):
    ○ A affects/binds/regulates/interacts with B
    ○ B affects/binds/regulates/interacts with C
    ○ A and C are not explicitly linked in any article
    => There might be an undiscovered relationship between A and C

    Various reasons why A and C might not be connected, e.g. B is a rare


    term.
    Example 1: Literature Based Discovery
    A B
    Term and relation
    Papers on extraction Papers
    small-cell lung resulting from
    carcinoma List of terms term queries
    search engine

    C
    Recognise entities, Term and
    such as symptoms, relation
    Ranker
    Lists B and C conditions, therapeutic extraction
    with semantic agents (drugs) List of terms
    annotations

    Knowledge
    base

    Ranked list of
    hypotheses
    connecting A
    and C
    Medical
    researcher
    Example 1: Visualising the A-C
    hypotheses

    Reproduced with permission from FastVac and Public Health England.


    Example 2: modeling research arguments
    using citation contexts
    Example 2: modeling research arguments
    using citation contexts
    ● Label the citation network according to the context of how
    they are cited
    ● Pioneering work by Simone Teufel in this area
    Green: Contrastive or
    comparative statements
    Pink: Statements about
    the aim of the paper

    Source: Simone Teufel


    (1999) Argumentative
    Zoning: Information
    Extraction from Scientific
    Text, Phd Thesis,
    University of Edinburgh.
    Example 3: exploratory search over
    research papers

    Source: Herrmannova & Knoth (2012) Visual Search for Supporting Content Exploration in Large Document
    Collections, DLib 18(8).
    How can I get started?
    • Get the data
    • Design/build your workflow
    • Select a framework, tools, services to be used in implementing the
    workflow
    • Understand how to evaluate the performance of each component
    Full-text Open Access article sources
    • Subject repositories:
    • arXiv bulk data

    • PubMed OA Subset

    • Institutional repositories:
    • ~3k across the world, see Directory of Open Access Repositories

    • Open Access journals:


    • >10k OA journals, see Directory of Open Access Journals (DOAJ)

    • Open Access subsets from publishers:


    • Elsevier OA STM Corpus
    • SpringerOpen

    • Aggregators:
    • CORE (API, Data dumps)
    Other full-text sources
    Typically for non-commercial research only.

    • Publisher full-text APIs:


    • Elsevier’s Text and Data Mining API:
    • (subscribed content on ScienceDirect)
    • Domain specific corpuses:
    • JSTOR Data For Research
    • Full-text content from multiple publishers:
    • CrossRef Text and Data Mining API (full-texts)
    • Mendeley API (abstracts)
    • Other useful scholarly resources:
    • Microsoft Academic Graph (MAG)
    • Clinical records
    • i2b2 NLP Challenge Data Sets
    Building a text mining application
    Extraction and Sentence
    Tokenisation
    Papers text cleaning splitting

    Stemming, Part of Speech


    lemmatisation Parsing
    Tagging (POS)

    Named Entity
    Relation extraction
    Recognition (NER)

    Semantic Analysis
    search, Knowledge and
    reasoning, ... base visualisation
    Example: Part of speech tagging
    [Acute]JJ [lymphoblastic]JJ [leukemia]NN ([ALL]NN) [leads]VB [to]IN [an]DT
    [accumulation]NN [of]IN [immature]JJ [lymphoid]JJ [cells]NN [into]IN [the]DT [bone]NN
    [marrow]NN, [blood]NN [and]CC [other]JJ [organs]NN.

    [A]DT [young]JJ [patient]NN [was]VB [treated]VB [unconventionally]RB [for]IN


    [Philadelphia]NN [positive]NN [ALL]NN [and]CC [Mucormycosis]NN [with]IN
    [Amphotericin]NN [B]NN.

    [The]DT [patient]NN [achieved]VB [a]DT [disease-free]JJ [survival]NN [of]IN [12]CD


    [months]NN [with]IN [good]JJ [quality]NN [of]IN [life]NN.
    Example: Parsing
    [Acute lymphoblastic leukemia]NP ([ALL]NP) [leads to]VP [an]DT [accumulation]NN
    [of]IN [immature lymphoid cells]NP [into]IN [the]DT [bone marrow]NP, [blood]NP
    [and]CC [other organs]NP.

    [A]DT [young patient]NP [was treated unconventionally for]VP [Philadelphia


    positive ALL]NP [and]CC [Mucormycosis]NP [with]IN [Amphotericin B]NP.

    [The]DT [patient]NP [achieved]VP [a]DT [disease-free survival]NP [of]IN [12 months]

    NP
    [with]IN [good]JJ [quality of life]PP.
    Example: Named entity recognition
    [Acute lymphoblastic leukemia]Disease ([ALL]Disease) [RESULT]VP [accumulation]

    NN
    [of]IN [immature lymphoid cells]AnatomicalSite [into]IN [the]DT [bone marrow]

    AnatomicalSite
    , [blood]AnatomicalSite [and]CC [other organs]AnatomicalSite.

    [A]DT [young patient]Person [TREAT]VP [Philadelphia positive ALL]Disease [and]CC


    [Mucormycosis]Disease [with]IN [Amphotericin B]Treatment.

    [The]DT [patient]Person [RESULT]VP [disease-free survival]Outcome [of]IN [12 months]

    Duration
    [with]IN [good quality of life]Outcome.
    Example: Named entity recognition

    ● Here, ALL is automatically expanded to its full form: acute lymphoblastic leukemia
    ● ALL is automatically labelled as a DiseaseOrSyndrome. as the initial mention was also labelled
    as DiseaseOrSyndrome
    Example: Named entity recognition

    ● Philadelphia on its own can be labelled as a Location


    ● In this context it is part of a longer noun phrase ‘Philadelphia positive ALL’, of which ALL is the
    Disease acute lymphoblastic leukemia
    ● ‘Philadelphia positive’ is also a term in a domain-specific dictionary
    ● Which label is chosen can be determined by a rule, either handwritten or learnt by a machine
    learning algorithm
    Frameworks
    ● General Architecture for Text Engineering (GATE)
    ● Apache UIMA
    ● Natural Language Toolkit (NLTK)
    ● OpenNLP
    Tools and Services
    ● Open Calais: general purpose tagging of people, places, companies, facts, events and
    relationships
    ● AlchemyAPI: similar to OpenCalais
    ● National Centre for Text Mining: terms, entities, semantic search
    ○ Cafetiere: text mine your own documents online
    ○ EventMine: relations between terms: protein binding, synthesis inhibition
    ● ContentMine: Fact extraction
    ● ReVerb: open-domain relation extraction
    ● XIP Parser: linguistic analysis
    ● Linguamatics
    ● Metadata and citation extraction:
    ○ ParsCit
    ○ Grobid
    ○ Cermine
    Open source demos

    ● Mimir
    ○ analysing, comparing and contrasting research findings
    ● EEXCESS
    ○ Recommend related resources
    ○ Narrative paths between resources
    Existing challenges
    ● Harmonising metadata formats across publishers, repositories, etc.
    ● Agreeing on standards/ontologies/formats used to share the outputs of
    research publications text-mining tools and all their components.
    ● Integration of text-mining tools with content providers’ systems
    ● Building and maintaining text-mining web-services for research (building
    blocks)
    ● Promoting and adopting end-user tools utilising text-mining in
    researchers’ daily workflows
    ● Building gold standards for various text-mining tasks and sharing them
    across researchers (issue of credit)
    Events and initiatives
    ● Events
    ○ International Workshop on Mining Scientific Publications (WOSP)
    ○ Joint Conference on Digital Libraries (JCDL)
    ○ ACL, COLING, IJC-NLP, CoNLL, LREC, etc.

    ● Projects
    ○ OpenMinTeD (EC funded)
    ○ EEXCESS
    ○ OpenAIRE
    Thanks for listening ...
    datascience@mendeley.com

    You might also like