DOCSLIB.ORG

Analysis of Link Grammar on Biomedical Dependency Corpus

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Analysis of Link Grammar on Biomedical Dependency Corpus Analysis of Link Grammar on Biomedical Dependency Corpus Targeted at Protein-Protein Interactions Sampo Pyysalo, Filip Ginter, Tapio Pahikkala, Jeppe Koivula Jorma Boberg, Jouni JÄarvinen, Tapio Salakoski MediCel Ltd., Turku Centre for Computer Science (TUCS) Haartmaninkatu 8 and Dept. of computer science, University of Turku 00290 Helsinki, Finland, LemminkÄaisenkatu 14A [email protected] 20520 Turku, Finland, ¯rst name.last [email protected].¯ Abstract parser with the UMLS Specialist2 lexicon, and Ding et al. (2003) perform a basic evaluation In this paper, we present an evaluation of the of LG performance on biomedical text. As both Link Grammar parser on a corpus consisting papers suggest, LG will require modi¯cations in of sentences describing protein-protein interac- order to provide a correct analysis of grammat- tions. We introduce the notion of an interac- ical phenomena that are rare in general English tion subgraph, which is the subgraph of a de- text, but common in biomedical language. Im- pendency graph expressing a protein-protein in- plementing such modi¯cations is a major e®ort teraction. We measure the performance of the that requires a careful analysis of the perfor- parser for recovery of dependencies, fully correct mance of the LG parser to identify the most linkages and interaction subgraphs. We analyze common causes of parsing failures and to target the causes of parser failure and report speci¯c modi¯cation e®orts. causes of error, and identify potential modi¯ca- While Szolovits (2003) does not attempt to tions to the grammar to address the identi¯ed evaluate parser performance at all and Ding issues. We also report and discuss the e®ect of et al. (2003) provide only an informal evalua- an extension to the dictionary of the parser. tion on manually simpli¯ed sentences, we focus on a more formal evaluation of the LG parser. 1 Introduction For the purpose of this study and also for sub- The challenges of processing the vast amounts of sequent research of biomedical information ex- biomedical publications available in databases traction with the LG parser, we have developed such as MEDLINE have recently attracted a a hand-annotated corpus consisting of unmod- considerable interest in the Natural Language i¯ed sentences from publications. We use this Processing (NLP) research community. The corpus to evaluate the performance of the LG task of information extraction, commonly tar- parser and to identify problems and potential geting entity relationships, such as protein- improvements to the grammar and parser. protein interactions, is an often studied prob- lem to which various NLP methods have been 2 Link Grammar and parser applied, ranging from keyword-based methods The Link Grammar and its parser represent an (see, e.g., Ginter et al. (2004)) to full syntactic implementation of a dependency-based compu- analysis as employed, for example, by Craven tational grammar. The result of LG analysis for and Kumlien (1999), Temkin and Gilder (2003) a sentence is a labeled undirected simple graph, and Daraselia et al. (2004). whose nodes represent the words of the sentence In this paper, we focus on the syntactic anal- and whose edges and their labels express the ysis component of an information extraction grammatical relationships between the words. system targeted to ¯nd protein-protein inter- In LG terminology, the graph is called a link- actions from the dependency output produced age links 1 , and its edges are called . The linkage by the Link Grammar (LG) parser of Sleator must be planar (i.e., links must not cross) when and Temperley (1991). Two recent papers study drawn above the words of the sentence, and the LG in the context of biomedical NLP. The work labels of the links must satisfy the linking con- by Szolovits (2003) proposes a fully automated straints speci¯ed for each word in the grammar. method to extend the dictionary of the LG A connected linkage is termed complete. 1http://www.link.cs.cmu.edu/link/ 2http://www.nlm.nih.gov/research/umls/ 15 findings suggest that PIP2 binds to proteins such as profilin Figure 1: Annotation example. The interaction of two proteins, PIP2 and pro¯lin, is stated by the words binds to. The links joining these words form the interaction subgraph (drawn with solid lines). Due to the structural ambiguity of natural names of at least two proteins that are known language, several linkages can typically be con- to interact. A domain expert annotated these structed for an input sentence. In such cases, sentences for protein names and for words stat- the LG parser enumerates all linkages allowed ing their interactions. Of these sentences, 1114 by the grammar. A post-processing step is then described at least one protein-protein interac- employed to enforce a number of additional con- tion. straints. The number of linkages for some sen- Thereafter, we performed a dependency anal- tences can be very high, making post-processing ysis and produced annotation of dependencies. and storage prohibitively expensive. This prob- To minimize the amount of mistakes, each sen- lem is addressed in the LG parser by de¯ning tence was independently annotated by two an- kmax, the maximal number of linkages to be notators and di®erences were then resolved by post-processed. If the parsing algorithm pro- discussion. The assigned dependency structure duces more than kmax linkages, the output is was produced according to the LG linkage con- reduced to kmax linkages by random sampling. ventions. Link types were not included in the The linkages are then ordered from best to worst annotation, and no cycles were introduced in using heuristic goodness criteria. the dependency graphs. All ambiguities where In order to be usable in practice, a parser is the LG parser is capable of at least enumerat- typically required to provide a partial analysis ing all alternatives (such as prepositional phrase of a sentence for which it cannot construct a full attachment) were enforced in the annotation. analysis. If the LG parser cannot construct a A random sample consisting of 300 sentences, complete linkage for a sentence, the connected- including 28 publication titles, has so far been ness requirement is relaxed so that some words fully annotated, giving 7098 word-to-word de- do not belong to the linkage at all. The LG pendencies. This set of sentences is the corpus parser is also time-limited. If the full set of we refer to in the following sections. linkages cannot be constructed in a given time An information extraction system targeted tmax, the parser enters a panic mode, in which it at protein-protein interactions and their types performs an e±cient but considerably restricted needs to identify three constituents that express parse, resulting in reduced performance. The an interaction in a sentence: the proteins in- parameters tmax and kmax set the trade-o® be- volved and the word or phrase that states their tween the qualitative performance and the re- interaction and suggests the type of this inter- source e±ciency of the parser. action. To extract this information from a LG linkage, the links connecting these items must 3 Corpus annotation and interaction be recovered correctly by the parser. The fol- subgraphs lowing de¯nition formalizes this notion. To compile a corpus of sentences describing De¯nition 1 (Interaction subgraph) The protein-protein interactions, we ¯rst selected interaction subgraph for an interaction between pairs of proteins that are known to interact two proteins A and B in a linkage L is the 3 from the Database of Interacting Proteins . We minimal connected subgraph of L that contains entered these pairs as search terms into the A, B, and the word or phrase that states their PubMed retrieval system. We then split the interaction. publication abstracts returned by the searches into sentences and included titles. These were The recovery of a connected component con- again searched for the protein pairs. This gave taining the protein names and the interaction us a set of 1927 sentences that contain the word is not su±cient: by the de¯nition of a complete linkage, such a component is always 3http://dip.doe-mbi.ucla.edu/ present. Consequently, the exact set of links 16 that forms the interaction subgraph must be re- time tmax for producing a normal parse was ex- covered. hausted and the parser entered panic mode, (2) For each interaction stated in a sentence, sentences where linkages were sampled because the corpus annotation speci¯es the proteins in- more than kmax linkages were produced, and volved and the interaction word. The interac- (3) stable sentences for which neither of these tion subgraph for each interaction can thus be occurred. A full analysis of all linkages that extracted automatically from the corpus. Be- the grammar allows is only possible for stable cause the corpus does not contain cyclic depen- sentences. For sentences in the other two cat- dencies, the interaction subgraphs are unique. egories, random e®ects may a®ect the results: 366 interaction subgraphs were identi¯ed from sentences for which more than kmax linkages are the corpus, one for each described interaction. produced are subject to randomness in sam- The interaction subgraphs can be partially over- pling, and sentences where the parser enters lapping, because a single link can be part of panic mode were always subject to subsequent more than one interaction subgraph. Figure 1 sampling in our experiments. shows an example of an annotated text frag- ment. 5 Evaluation To evaluate the ability of the LG parser to pro- 4 Evaluation criteria duce correct linkages, we increased the number of stable sentences by setting the tmax param- We evaluated the performance of the LG parser eter to 10 minutes and the kmax parameter to according to the following three quantitative cri- 10000 instead of using the defaults tmax = 30 teria: seconds and kmax = 1000.
Load more

Recommended publications
  • A Protein Interaction Extraction Systemusing a Link Grammar Parser from Biomedical Abstracts
    World Academy of Science, Engineering and Technology International Journal of Biomedical and Biological Engineering Vol:1, No:5, 2007 PIELG: A Protein Interaction Extraction System using a Link Grammar Parser from Biomedical Abstracts Rania A. Abul Seoud, Nahed H. Solouma, Abou-Baker M. Youssef, and Yasser M. Kadah, Senior Member, IEEE failure. Applications that repair or replace portions of or Abstract—Due to the ever growing amount of publications about whole living tissues (e.g., bone, dentine, or bladder) using protein-protein interactions, information extraction from text is living cells is named Tissue Engineering (TE). For example, increasingly recognized as one of crucial technologies in dentine formation is the process of regenerating dental tissues bioinformatics. This paper presents a Protein Interaction Extraction by tissue engineering principles and technology. Dentine System using a Link Grammar Parser from biomedical abstracts formation is governed by biological mediators or growth (PIELG). PIELG uses linkage given by the Link Grammar Parser to start a case based analysis of contents of various syntactic roles as factors (protein) and interactions amongst different proteins. well as their linguistically significant and meaningful combinations. Dentine formation needs the support of continuous updated The system uses phrasal-prepositional verbs patterns to overcome information about protein-protein interactions. preposition combinations problems. The recall and precision are Researches in the last decade have resulted in the 74.4% and 62.65%, respectively. Experimental evaluations with two production of a large amount of information about protein other state-of-the-art extraction systems indicate that PIELG system functions involved in dentine formation process. That achieves better performance.
    [Show full text]
  • Implementing a Portable Clinical NLP System with a Common Data Model: a Lisp Perspective
    Implementing a portable clinical NLP system with a common data model: a Lisp perspective The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters. Citation Luo, Yuan, and Peter Szolovits, "Implementing a portable clinical NLP system with a common data model: a Lisp perspective." Proceedings, 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM 2018), December 3-6, 2018, Madrid, Spain (Piscataway, N.J.: IEEE, 2018): doi 10.1109/BIBM.2018.8621521 ©2018 Author(s) As Published 10.1109/BIBM.2018.8621521 Publisher Institute of Electrical and Electronics Engineers (IEEE) Version Original manuscript Citable link https://hdl.handle.net/1721.1/124439 Terms of Use Creative Commons Attribution-Noncommercial-Share Alike Detailed Terms http://creativecommons.org/licenses/by-nc-sa/4.0/ Implementing a Portable Clinical NLP System with a Common Data Model – a Lisp Perspective Yuan Luo* Peter Szolovits* Dept. of Preventive Medicine CSAIL Northwestern University MIT Chicago, USA Cambridge, USA [email protected] [email protected] Abstract— This paper presents a Lisp architecture for a annotations, often in idiosyncratic representations. This makes portable NLP system, termed LAPNLP, for processing clinical it quite difficult to chain together sequences of operations. Alt- notes. LAPNLP integrates multiple standard, customized and hough several recent projects have achieved reasonable in-house developed NLP tools. Our system facilitates portability success in analyzing certain types of clinical narratives [3-6], across different institutions and data systems by incorporating an enriched Common Data Model (CDM) to standardize neces- efforts towards a common data model (CDM) to ensure port- sary data elements.
    [Show full text]
  • Chapter 14: Dependency Parsing
    Speech and Language Processing. Daniel Jurafsky & James H. Martin. Copyright © 2021. All rights reserved. Draft of September 21, 2021. CHAPTER 14 Dependency Parsing The focus of the two previous chapters has been on context-free grammars and constituent-based representations. Here we present another important family of dependency grammars grammar formalisms called dependency grammars. In dependency formalisms, phrasal constituents and phrase-structure rules do not play a direct role. Instead, the syntactic structure of a sentence is described solely in terms of directed binary grammatical relations between the words, as in the following dependency parse: root dobj det nmod (14.1) nsubj nmod case I prefer the morning flight through Denver Relations among the words are illustrated above the sentence with directed, labeled typed dependency arcs from heads to dependents. We call this a typed dependency structure because the labels are drawn from a fixed inventory of grammatical relations. A root node explicitly marks the root of the tree, the head of the entire structure. Figure 14.1 shows the same dependency analysis as a tree alongside its corre- sponding phrase-structure analysis of the kind given in Chapter 12. Note the ab- sence of nodes corresponding to phrasal constituents or lexical categories in the dependency parse; the internal structure of the dependency parse consists solely of directed relations between lexical items in the sentence. These head-dependent re- lationships directly encode important information that is often buried in the more complex phrase-structure parses. For example, the arguments to the verb prefer are directly linked to it in the dependency structure, while their connection to the main verb is more distant in the phrase-structure tree.
    [Show full text]
  • Development of a Persian Syntactic Dependency Treebank
    Development of a Persian Syntactic Dependency Treebank Mohammad Sadegh Rasooli Manouchehr Kouhestani Amirsaeid Moloodi Department of Computer Science Department of Linguistics Department of Linguistics Columbia University Tarbiat Modares University University of Tehran New York, NY Tehran, Iran Tehran, Iran [email protected] [email protected] [email protected] Abstract tions in tasks such as machine translation. Depen- dency treebanks are collections of sentences with This paper describes the annotation process their corresponding dependency trees. In the last and linguistic properties of the Persian syn- decade, many dependency treebanks have been de- tactic dependency treebank. The treebank veloped for a large number of languages. There are consists of approximately 30,000 sentences at least 29 languages for which at least one depen- annotated with syntactic roles in addition to morpho-syntactic features. One of the unique dency treebank is available (Zeman et al., 2012). features of this treebank is that there are al- Dependency trees are much more similar to the hu- most 4800 distinct verb lemmas in its sen- man understanding of language and can easily rep- tences making it a valuable resource for ed- resent the free word-order nature of syntactic roles ucational goals. The treebank is constructed in sentences (Kubler¨ et al., 2009). with a bootstrapping approach by means of available tagging and parsing tools and man- Persian is a language with about 110 million ually correcting the annotations. The data is speakers all over the world (Windfuhr, 2009), yet in splitted into standard train, development and terms of the availability of teaching materials and test set in the CoNLL dependency format and annotated data for text processing, it is undoubt- is freely available to researchers.
    [Show full text]
  • Downloads." the Open Information Security Foundation
    Performance Testing Suricata The Effect of Configuration Variables On Offline Suricata Performance A Project Completed for CS 6266 Under Jonathon T. Giffin, Assistant Professor, Georgia Institute of Technology by Winston H Messer Project Advisor: Matt Jonkman, President, Open Information Security Foundation December 2011 Messer ii Abstract The Suricata IDS/IPS engine, a viable alternative to Snort, has a multitude of potential configurations. A simplified automated testing system was devised for the purpose of performance testing Suricata in an offline environment. Of the available configuration variables, seventeen were analyzed independently by testing in fifty-six configurations. Of these, three variables were found to have a statistically significant effect on performance: Detect Engine Profile, Multi Pattern Algorithm, and CPU affinity. Acknowledgements In writing the final report on this endeavor, I would like to start by thanking four people who made this project possible: Matt Jonkman, President, Open Information Security Foundation: For allowing me the opportunity to carry out this project under his supervision. Victor Julien, Lead Programmer, Open Information Security Foundation and Anne-Fleur Koolstra, Documentation Specialist, Open Information Security Foundation: For their willingness to share their wisdom and experience of Suricata via email for the past four months. John M. Weathersby, Jr., Executive Director, Open Source Software Institute: For allowing me the use of Institute equipment for the creation of a suitable testing
    [Show full text]
  • Lexicalization and Grammar Development
    University of Pennsylvania ScholarlyCommons IRCS Technical Reports Series Institute for Research in Cognitive Science April 1995 Lexicalization and Grammar Development B. Srinivas University of Pennsylvania Dania Egedi University of Pennsylvania Christine D. Doran University of Pennsylvania Tilman Becker University of Pennsylvania Follow this and additional works at: https://repository.upenn.edu/ircs_reports Srinivas, B.; Egedi, Dania; Doran, Christine D.; and Becker, Tilman, "Lexicalization and Grammar Development" (1995). IRCS Technical Reports Series. 138. https://repository.upenn.edu/ircs_reports/138 University of Pennsylvania Institute for Research in Cognitive Science Technical Report No. IRCS-95-12. This paper is posted at ScholarlyCommons. https://repository.upenn.edu/ircs_reports/138 For more information, please contact [email protected]. Lexicalization and Grammar Development Abstract In this paper we present a fully lexicalized grammar formalism as a particularly attractive framework for the specification of natural language grammars. We discuss in detail Feature-based, Lexicalized Tree Adjoining Grammars (FB-LTAGs), a representative of the class of lexicalized grammars. We illustrate the advantages of lexicalized grammars in various contexts of natural language processing, ranging from wide-coverage grammar development to parsing and machine translation. We also present a method for compact and efficientepr r esentation of lexicalized trees. In diesem Beitrag präsentieren wir einen völlig lexikalisierten Grammatikformalismus als eine besonders geeignete Basis für die Spezifikation onv Grammatiken für natürliche Sprachen. Wir stellen feature- basierte, lexikalisierte Tree Adjoining Grammars (FB-LTAGs) vor, ein Vertreter der Klasse der lexikalisierten Grammatiken. Wir führen die Vorteile von lexikalisierten Grammatiken in verschiedenen Bereichen der maschinellen Sprachverarbeitung aus; von der Entwicklung von Grammatiken für weite Sprachbereiche über Parsing bis zu maschineller Übersetzung.
    [Show full text]
  • Parsing Aligned Parallel Corpus by Projecting Syntactic Relations from Annotated Source Corpus
    Parsing Aligned Parallel Corpus by Projecting Syntactic Relations from Annotated Source Corpus Shailly Goyal Niladri Chatterjee Department of Mathematics Indian Institute of Technology Delhi Hauz Khas, New Delhi - 110 016, India {shailly goyal, niladri iitd}@yahoo.com Abstract approaches either use examples from the same lan- guage, e.g., (Bod et al., 2003; Streiter, 2002), or Example-based parsing has already been they try to imitate the parse of a given sentence proposed in literature. In particular, at- using the parse of the corresponding sentence in tempts are being made to develop tech- some other language (Hwa et al., 2005; Yarowsky niques for language pairs where the source and Ngai, 2001). In particular, Hwa et al. (2005) and target languages are different, e.g. have proposed a scheme called direct projection Direct Projection Algorithm (Hwa et al., algorithm (DPA) which assumes that the relation 2005). This enables one to develop parsed between two words in the source language sen- corpus for target languages having fewer tence is preserved across the corresponding words linguistic tools with the help of a resource- in the parallel target language. This is called Di- rich source language. The DPA algo- rect Correspondence Assumption (DCA). rithm works on the assumption of Di- However, with respect to Indian languages we rect Correspondence which simply means observed that the DCA does not hold good all the that the relation between two words of time. In order to overcome the difficulty, in this the source language sentence can be pro- work, we propose an algorithm based on a vari- jected directly between the correspond- ation of the DCA, which we call pseudo Direct ing words of the parallel target language Correspondence Assumption (pDCA).
    [Show full text]
  • Dependency-Length Minimization in Natural and Artificial Languages*
    Journal of Quantitative Linguistics 2008, Volume 15, Number 3, pp. 256 – 282 DOI: 10.1080/09296170802159512 Dependency-Length Minimization in Natural and Artificial Languages* David Temperley Eastman School of Music at the University of Rochester ABSTRACT A wide range of evidence points to a preference for syntactic structures in which dependencies are short. Here we examine the question: what kinds of dependency configurations minimize dependency length? We consider two well-established principles of dependency-length minimization; that dependencies should be consistently right-branching or left-branching, and that shorter dependent phrases should be closer to the head. We also add a third, novel, principle; that some ‘‘opposite-branching’’ of one-word phrases is desirable. In a series of computational experiments, using unordered dependency trees gathered from written English, we examine the effect of these three principles on dependency length, and show that all three contribute significantly to dependency-length reduction. Finally, we present what appears to be the optimal ‘‘grammar’’ for dependency-length minimization. INTRODUCTION A wide range of evidence from diverse areas of linguistic research points to a preference for syntactic structures in which dependencies between related words are short.1 This preference is reflected in numerous psycholinguistic phenomena: structures with long depen- dencies appear to cause increased complexity in comprehen- sion and tend to be avoided in production. The preference for shorter dependencies has also been cited as an explanation for certain *Address correspondence to: David Temperley, Eastman School of Music, 26 Gibbs Street, Rochester, NY 14604. Tel: 585-274-1557. E-mail: [email protected] 1Thanks to Daniel Gildea for valuable comments on a draft of this paper.
    [Show full text]
  • Natural Language Processing (NLP): Written and Oral Involves Processing of Written Text Using Computer Models at Lexical, Syntactic, and Semantic Level
    Natural Language Processing (NLP): written and oral Involves processing of written text using computer models at lexical, syntactic, and semantic level. Introduction: 1. Understanding (of language) – map text/speech to immediately useful form. 2. Generation (of language) Sentence Analysis Phases: 1. Morphological Analysis: extracts root word from declined/inflection form of word after removing suffices and prefixes. 2. Syntactic Analysis: builds a structural description of a sentence using MAP (Morphological Analysis Process) based on grammatical rules, called Parsing. 3. Semantic Analysis: Creates a semantic structure by ascribing a literal meaning to sentence using parse structure obtained in syntactic phase. It maps individual words into corresponding objects in the knowledge base. Focus is on creation of target representation of the meaning of a sentence. 4. Pragmatic Analysis: to establish meaning of a sentence in different contexts. 5. Discourse Analysis: Interpretation based on belief during conversation. Grammars and Parsers: Parsing: Process of analyzing an input sequence in order to determine its structure with respect to a given grammar. Types of Parsing: 1. Rule-based parsing: syntactic structure of language is provided in the form of linguistic rules which can be coded as production rules that are similar to context-free rules. 1. Top-down parsing: Start with start symbol and apply grammar rules in the forward direction until the terminal symbols of the parse tree correspond to the words in the sentence. 2. Bottom-up parsing: Start with the words in the sentence in the sentence and apply grammar rules in the backward direction until a single tree is produced whose root matches with the start symbol.
    [Show full text]
  • Extraction of Protein Interaction Information from Unstructured Text Using a Link Grammar Parser Rania A
    Extraction of Protein Interaction Information from Unstructured Text Using a link Grammar Parser Rania A. Abul Seoud1, Nahed H. Solouma2, Abou-Bakr M. Youssef3, Yasser M. Kadah3 1,3Misr University for Science and Technology, 6 October City, Egypt 2Laser Institute, Cairo University, Giza, Egypt 3Biomedical Engineering Department, Faculty of Engineering, Cairo University, Giza, Egypt E-mail: [email protected] Abstract- As research into disease pathology and cellular number of groups reported application of pattern-matching- function continues to generate vast amounts of data, pertaining based systems for protein-function information extraction. to protein and gene interactions, there is a critical need to The shortcoming of such systems is their inability to process capture these results in structured formats allowing for correctly anything other than short, straightforward computational analysis. Although many efforts have been made statements, which are quite rare in information-saturated to create databases that store this information in computer readable form, populating these sources largely requires a MEDLINE and PubMED abstracts. Several attempts have manual process of interpreting and extracting interaction been made to utilize shallow-parsing techniques for the task relationships from the biological research literature. Being able of biological information. Shallow parsers perform partial to efficiently and accurately automate the extraction of decomposition of a sentence structure. In some cases, interactions from unstructured text, would improve the content shallow-parsers are used in combination with various of these databases, and provide a method for managing the heuristic and statistical methods. Information extraction continued growth of new literature being published. Hence, it is systems based on the full-sentence parsing approach deal important to have a fully automated extraction system for the with the structure of an entire sentence.
    [Show full text]
  • Wanderlust: Extracting Semantic Relations from Natural Language Text Using Dependency Grammar Patterns
    Wanderlust: Extracting Semantic Relations from Natural Language Text Using Dependency Grammar Patterns Alan Akbik Jürgen Broß Freie Universität Berlin Freie Universität Berlin Institute of Computer Science Institute of Computer Science 14195 Berlin, Germany 14195 Berlin, Germany [email protected] [email protected] ABSTRACT SUMO [20] have been manually constructed in a laborious A great share of applications in modern information tech- and expensive process. While such a procedure offers very nology can benefit from large coverage, machine accessible high precision, it naturally does not scale well. High cover- knowledge bases. However, the bigger part of todays knowl- age and up-to-dateness are difficult to be achieved this way. edge is provided in the form of unstructured data, mostly Inspired by success stories such as Wikipedia, recent ap- plain text. As an initial step to exploit such data, we present proaches pursue the idea of building large-scale structured Wanderlust, an algorithm that automatically extracts se- knowledge bases by leveraging the collaborative power of mantic relations from natural language text. The procedure thousands of volunteers [26, 22]. We believe this to be a uses deep linguistic patterns that are defined over the depen- very promising way. However, similar to other Web 2.0-style dency grammar of sentences. Due to its linguistic nature, applications, such an approach depends heavily on a critical the method performs in an unsupervised fashion and is not mass of users. In fact, this approach entails a circular de- restricted to any specific type of semantic relation. The ap- pendency: A critical mass of users can only be attracted if plicability of the proposed approach is examined in a case enough structured data has been collected so that people can study, in which it is put to the task of generating a seman- start developing compelling applications.
    [Show full text]
  • Interpretable Natural Language Segmentation Based on Link Grammar
    Interpretable Natural Language Segmentation Based on Link Grammar Vignav Ramesh Anton Kolonin SingularityNET, Saratoga H.S. Novosibirsk State University Saratoga, CA, USA Novosibirsk, Russia November 15, 2020 November 15, 2020 1 / 27 Overview I Introduction • Natural Language Segmentation • Sources of Unsegmented Text • Motivations • Link Grammar I Methodology • Loader • Segment I Results I Conclusion • Applications • Future Work November 15, 2020 2 / 27 Introduction November 15, 2020 3 / 27 Natural Language Segmentation I Process of dividing written text into meaningful units I Focus on the sub-problem of sentence segmentation • Dividing a string of natural language (NL) text into its component sentences that can then be used to build a parse tree • Proposed architecture solely relies on grammatical relationships between tokens I Generalizes to languages other than English or Russian November 15, 2020 4 / 27 Sources of Unsegmented Text 1. Text output from speech-to-text (STT) recognition engines • STT engines transcribe audio content and return sequences of tokens without delimeters • Often used to preprocess the inputs of voice-activated question answering or chatbot pipelines (among various other scenarios involving the transcription of audio without recorded punctuation) 2. Crawled web pages • Returns raw HTML and CSS output that often does not contain proper punctuation • HTML and CSS tags are neither accurate nor consistent sentence boundaries November 15, 2020 5 / 27 Motivations November 15, 2020 6 / 27 Interpretable Natural Language
    [Show full text]