pith. sign in

arxiv: 2605.23710 · v1 · pith:YV3FFGVBnew · submitted 2026-05-22 · 💻 cs.CL

A graph-based analysis of semantic types and coercion in contextualized word embeddings

Long Chen , Deniz Ekin Yavas This is my paper

Pith reviewed 2026-05-25 04:17 UTC · model grok-4.3

classification 💻 cs.CL
keywords semantic typescoercioncontextualized embeddingsgraph analysisneighbor type probabilityneighbor type entropyBERTsense-enhanced embeddings
0
0 comments X

The pith

Sense-enhanced embeddings allow graphs to better reflect semantic types and distinguish coercion from matching contexts via neighbor metrics.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that semantic type information for nouns is more clearly encoded in sense-enhanced embeddings than in standard BERT embeddings when viewed through graph neighborhoods. By selecting nouns from ten semantic types and annotating sentences for whether the context matches the type, involves coercion, other mismatch, or is unrestricted, the authors build similarity graphs and introduce two metrics to measure how neighbors share types. A sympathetic reader would care because coercion is a common way language stretches meanings, and if embeddings capture the difference, it opens ways to probe linguistic knowledge in models without task-specific training. The results indicate that sense-enhanced versions produce neighborhoods where type distributions are more informative, allowing the metrics to separate the categories. This approach treats embeddings as graphs to reveal hidden structure in how context and lexicon interact.

Core claim

Graphs constructed with sense-enhanced embeddings reflect semantic type information better than those using BERT embeddings alone. The Neighbor Type Probability and Neighbor Type Entropy metrics applied to these graphs distinguish matching sentences from coercion and mismatch sentences in the annotated data.

What carries the argument

Graph neighborhoods of contextualized embeddings measured by Neighbor Type Probability (NTP), the probability a neighbor shares the noun's semantic type, and Neighbor Type Entropy (NTE), the entropy of type distribution in neighbors.

If this is right

  • Sense-enhanced embeddings capture type information more effectively in their similarity graphs.
  • Matching contexts show higher NTP and lower NTE than coercion or mismatch contexts.
  • The proposed metrics provide a quantitative way to analyze type consistency in embeddings.
  • Distinctions between coercion and other mismatches are detectable through neighborhood analysis.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the metrics reliably detect coercion, they could be applied to unannotated text to identify potential type shifts automatically.
  • This graph approach might generalize to analyzing other contextual effects like metaphor that also involve type mismatches.
  • Future work could test whether these metrics predict human judgments of sentence acceptability in type-shifting contexts.

Load-bearing premise

The manual annotation of corpus instances into type matching, coercion, other mismatch, and unrestricted categories accurately reflects the semantic phenomena without significant annotator bias or error.

What would settle it

Finding that a new set of annotations by independent annotators changes the category labels for many instances and eliminates the reported differences in NTP and NTE between categories would falsify the claim.

Figures

Figures reproduced from arXiv: 2605.23710 by Deniz Ekin Yavas, Long Chen.

Figure 1
Figure 1. Figure 1: A presumed type hierarchy of the selected [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Neighbor Type Ratio (NT P) for MATCHING sentences per semantic type, from the graphs based on BERT (Gb) (left) and sense-enhanced BERT embeddings (Gs) (right). The diagonal grids correspond to the average Lexical Neighbor Type Matching Ratio (NTMRL) for each type. Entity Abstract Static info state mood Motion activity process Concrete Inanimate artifact food location Animate animal human [PITH_FULL_IMAGE:… view at source ↗
Figure 3
Figure 3. Figure 3: A new type hierarchy induced from the NT P values of the semantic types. masked models, suggesting that the greater diver￾sity of possible contextual types in UNRESTRICTED sentences is reflected more clearly on Gmb and Gms. The NT E further captures the distinction be￾tween MATCHING and UNRESTRICTED sentences. Across all four graphs, the diversity of neighbor type distributions is significantly higher for … view at source ↗
read the original abstract

Semantic type mismatch between a noun and its context is central to coercion phenomena. This paper introduces a graph-based method to examine how lexical and contextual type information is reflected in word embeddings. We select nouns from ten semantic types, annotate corpus instances for type matching (matching vs. coercion vs. other mismatch vs. unrestricted), and construct graphs using BERT and sense-enhanced embeddings. Two metrics -- Neighbor Type Probability (NTP) and Neighbor Type Entropy (NTE) -- are proposed to analyze neighborhood type distributions. Results show that graphs constructed with sense-enhanced embeddings reflect semantic type information better, and matching and mismatch sentences can be distinguished through the proposed metrics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces a graph-based method for analyzing semantic type information and coercion in contextualized embeddings. It selects nouns from ten semantic types, manually annotates corpus instances into four categories (type matching, coercion, other mismatch, unrestricted), constructs graphs from BERT and sense-enhanced embeddings, and applies two new metrics—Neighbor Type Probability (NTP) and Neighbor Type Entropy (NTE)—to neighborhood type distributions. The central claim is that sense-enhanced embeddings yield graphs that better reflect semantic types and that the metrics distinguish matching from mismatch sentences.

Significance. If the empirical distinctions hold after addressing annotation reliability, the work supplies a quantitative graph-neighborhood probe for how embeddings encode semantic type and coercion, extending existing diagnostic approaches in lexical semantics. The independent definition of NTP and NTE (applied to externally annotated nodes) is a methodological strength that avoids circularity.

major comments (2)
  1. [Methods] Annotation procedure (Methods section): All reported NTP/NTE comparisons rest on the four-category manual labels, yet the manuscript supplies no inter-annotator agreement figures, adjudication protocol, or external validation against established coercion diagnostics. If label noise exceeds typical thresholds for subtle semantic judgments, the claimed separations between sense-enhanced and BERT graphs (and between matching and mismatch sentences) could be artifacts of annotation variance rather than embedding properties.
  2. [Results] Experimental reporting (Results section): The abstract and results claim metric-based distinctions, but no details are given on per-category sample sizes, statistical significance tests, confidence intervals, or controls for confounds such as sentence length or frequency effects in graph construction. This prevents assessment of whether the observed differences are robust.
minor comments (1)
  1. [Abstract] The abstract states results without referencing the specific tables or figures that support the NTP/NTE comparisons; adding explicit cross-references would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our graph-based analysis of semantic types and coercion. We address the major comments below and will revise the manuscript to strengthen the reporting of annotation reliability and experimental details.

read point-by-point responses

  1. Referee: [Methods] Annotation procedure (Methods section): All reported NTP/NTE comparisons rest on the four-category manual labels, yet the manuscript supplies no inter-annotator agreement figures, adjudication protocol, or external validation against established coercion diagnostics. If label noise exceeds typical thresholds for subtle semantic judgments, the claimed separations between sense-enhanced and BERT graphs (and between matching and mismatch sentences) could be artifacts of annotation variance rather than embedding properties.

    Authors: We agree that inter-annotator agreement (IAA) reporting is necessary for manual semantic annotations. The original annotations were performed by a single expert following detailed guidelines derived from coercion literature (e.g., Pustejovsky 1995), which is why IAA was not initially reported. In revision we will annotate a 20% subset with a second annotator, compute Cohen's kappa, describe the adjudication protocol, and add these details to the Methods section. We will also briefly relate our four categories to prior coercion diagnostics for external grounding. These changes will allow readers to assess label reliability directly. revision: yes

  2. Referee: [Results] Experimental reporting (Results section): The abstract and results claim metric-based distinctions, but no details are given on per-category sample sizes, statistical significance tests, confidence intervals, or controls for confounds such as sentence length or frequency effects in graph construction. This prevents assessment of whether the observed differences are robust.

    Authors: We acknowledge the need for fuller experimental reporting. The manuscript contains the overall annotation counts but omits the per-category breakdown and statistical tests. In the revision we will add a table listing sample sizes per semantic type and category, report two-tailed t-tests (or appropriate non-parametric tests) comparing NTP/NTE across matching vs. coercion vs. other mismatch, include 95% confidence intervals, and add a supplementary analysis controlling for sentence length and token frequency by reporting means per category and, if needed, balanced subsampling. These additions will appear in the revised Results section and will be accompanied by the raw counts. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper selects nouns from ten semantic types, manually annotates corpus instances into four categories (matching, coercion, other mismatch, unrestricted), constructs graphs from BERT and sense-enhanced embeddings, and applies independently defined metrics NTP and NTE to neighborhood type distributions. These steps form a linear chain from external linguistic labels to metric computation with no reduction by construction, no fitted parameters renamed as predictions, and no load-bearing self-citations. The annotations function as independent ground truth rather than outputs derived from the embeddings or metrics.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Insufficient information from abstract alone to identify free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5628 in / 1063 out tokens · 48945 ms · 2026-05-25T04:17:11.333424+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

300 extracted references · 300 canonical work pages

  1. [1]

    Direct Error Rate Minimization for Statistical Machine Translation

    Chung, Tagyoung and Galley, Michel. Direct Error Rate Minimization for Statistical Machine Translation. Proceedings of the Seventh Workshop on Statistical Machine Translation. 2012

    work page 2012

  2. [2]

    Optimization Strategies for Online Large-Margin Learning in Machine Translation

    Eidelman, Vladimir. Optimization Strategies for Online Large-Margin Learning in Machine Translation. Proceedings of the Seventh Workshop on Statistical Machine Translation. 2012

    work page 2012

  3. [3]

    Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  4. [4]

    Towards Distributed MCMC Inference in Probabilistic Knowledge Bases

    Niepert, Mathias and Meilicke, Christian and Stuckenschmidt, Heiner. Towards Distributed MCMC Inference in Probabilistic Knowledge Bases. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  5. [5]

    Collectively Representing Semi-Structured Data from the Web

    Dalvi, Bhavana and Cohen, William and Callan, Jamie. Collectively Representing Semi-Structured Data from the Web. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  6. [6]

    Unsupervised Content Discovery from Concise Summaries

    Saggion, Horacio. Unsupervised Content Discovery from Concise Summaries. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  7. [7]

    Automatic Evaluation of Relation Extraction Systems on Large-scale

    Bronzi, Mirko and Guo, Zhaochen and Mesquita, Filipe and Barbosa, Denilson and Merialdo, Paolo. Automatic Evaluation of Relation Extraction Systems on Large-scale. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  8. [8]

    Relabeling Distantly Supervised Training Data for Temporal Knowledge Base Population

    Tamang, Suzanne and Ji, Heng. Relabeling Distantly Supervised Training Data for Temporal Knowledge Base Population. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  9. [9]

    Web Based Collection and Comparison of Cognitive Properties in E nglish and C hinese

    Li, Bin and Chen, Jiajun and Zhang, Yingjie. Web Based Collection and Comparison of Cognitive Properties in E nglish and C hinese. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  10. [10]

    Population of a Knowledge Base for News Metadata from Unstructured Text and Web Data

    Stern, Rosa and Sagot, Beno \^i t. Population of a Knowledge Base for News Metadata from Unstructured Text and Web Data. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  11. [11]

    Real-time Population of Knowledge Bases: Opportunities and Challenges

    Nakashole, Ndapandula and Weikum, Gerhard. Real-time Population of Knowledge Bases: Opportunities and Challenges. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  12. [12]

    Adding Distributional Semantics to Knowledge Base Entities through Web-scale Entity Linking

    Gardner, Matt. Adding Distributional Semantics to Knowledge Base Entities through Web-scale Entity Linking. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  13. [13]

    K rake N : N-ary Facts in Open Information Extraction

    Akbik, Alan and L. K rake N : N-ary Facts in Open Information Extraction. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  14. [14]

    Structural Linguistics and Unsupervised Information Extraction

    Grishman, Ralph. Structural Linguistics and Unsupervised Information Extraction. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  15. [15]

    A Context-Aware Approach to Entity Linking

    Stoyanov, Veselin and Mayfield, James and Xu, Tan and Oard, Douglas and Lawrie, Dawn and Oates, Tim and Finin, Tim. A Context-Aware Approach to Entity Linking. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  16. [16]

    Evaluating the Quality of a Knowledge Base Populated from Text

    Mayfield, James and Finin, Tim. Evaluating the Quality of a Knowledge Base Populated from Text. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  17. [17]

    Constructing a Textual KB from a Biology T ext B ook

    Clark, Peter and Harrison, Phil and Balasubramanian, Niranjan and Etzioni, Oren. Constructing a Textual KB from a Biology T ext B ook. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  18. [18]

    Knowledge Extraction and Joint Inference Using Tractable M arkov L ogic

    Kiddon, Chlo \'e and Domingos, Pedro. Knowledge Extraction and Joint Inference Using Tractable M arkov L ogic. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  19. [19]

    Entity Linking at Web Scale

    Lin, Thomas and Mausam and Etzioni, Oren. Entity Linking at Web Scale. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  20. [20]

    Human-Machine Cooperation: Supporting User Corrections to Automatically Constructed KB s

    Wick, Michael and Schultz, Karl and McCallum, Andrew. Human-Machine Cooperation: Supporting User Corrections to Automatically Constructed KB s. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  21. [21]

    Annotated G igaword

    Napoles, Courtney and Gormley, Matthew and Van Durme, Benjamin. Annotated G igaword. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  22. [22]

    Rel-grams: A Probabilistic Model of Relations in Text

    Balasubramanian, Niranjan and Soderland, Stephen and Mausam and Etzioni, Oren. Rel-grams: A Probabilistic Model of Relations in Text. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  23. [23]

    Automatic Knowledge Base Construction using Probabilistic Extraction, Deductive Reasoning, and Human Feedback

    Wang, Daisy Zhe and Chen, Yang and Goldberg, Sean and Grant, Christan and Li, Kun. Automatic Knowledge Base Construction using Probabilistic Extraction, Deductive Reasoning, and Human Feedback. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  24. [24]

    M onte C arlo MCMC : Efficient Inference by Sampling Factors

    Singh, Sameer and Wick, Michael and McCallum, Andrew. M onte C arlo MCMC : Efficient Inference by Sampling Factors. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  25. [25]

    Probabilistic Databases of Universal Schema

    Yao, Limin and Riedel, Sebastian and McCallum, Andrew. Probabilistic Databases of Universal Schema. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  26. [26]

    Using Textual Patterns to Learn Expected Event Frequencies

    Gordon, Jonathan and Schubert, Lenhart. Using Textual Patterns to Learn Expected Event Frequencies. Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction ( AKBC - WEKEX ). 2012

    work page 2012

  27. [27]

    Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  28. [28]

    A free and open-source tool that reads movie subtitles aloud

    Ljungl. A free and open-source tool that reads movie subtitles aloud. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  29. [29]

    and Carson-Berndsen, Julie

    Sz \'e kely, \'E va and Ahmed, Zeeshan and Cabral, Jo \ a o P. and Carson-Berndsen, Julie. W ink T alk: a demonstration of a multimodal speech synthesis platform linking facial expressions to expressive synthetic voices. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  30. [30]

    Discourse-Based Modeling for AAC

    Mitchell, Margaret and Sproat, Richard. Discourse-Based Modeling for AAC. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  31. [31]

    Applying Prediction Techniques to Phoneme-based AAC Systems

    Trinh, Ha and Waller, Annalu and Vertanen, Keith and Kristensson, Per Ola and Hanson, Vicki L. Applying Prediction Techniques to Phoneme-based AAC Systems. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  32. [32]

    Non-Syntactic Word Prediction for AAC

    Wiegand, Karl and Patel, Rupal. Non-Syntactic Word Prediction for AAC. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  33. [33]

    Assisting Social Conversation between Persons with A lzheimer ' s Disease and their Conversational Partners

    Green, Nancy and Guinn, Curry and Smith, Ronnie. Assisting Social Conversation between Persons with A lzheimer ' s Disease and their Conversational Partners. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  34. [34]

    Communication strategies for a computerized caregiver for individuals with A lzheimer ' s disease

    Rudzicz, Frank and Wilson, Rozanne and Mihailidis, Alex and Rochon, Elizabeth and Leonard, Carol. Communication strategies for a computerized caregiver for individuals with A lzheimer ' s disease. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  35. [35]

    Generating Situated Assisting Utterances to Facilitate Tactile-Map Understanding: A Prototype System

    Lohmann, Kris and Eichhorn, Ole and Baumann, Timo. Generating Situated Assisting Utterances to Facilitate Tactile-Map Understanding: A Prototype System. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  36. [36]

    Learning a Vector-Based Model of A merican S ign L anguage Inflecting Verbs from Motion-Capture Data

    Lu, Pengfei and Huenerfauth, Matt. Learning a Vector-Based Model of A merican S ign L anguage Inflecting Verbs from Motion-Capture Data. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  37. [37]

    A Hybrid System for S panish Text Simplification

    Bott, Stefan and Saggion, Horacio and Figueroa, David. A Hybrid System for S panish Text Simplification. Proceedings of the Third Workshop on Speech and Language Processing for Assistive Technologies. 2012

    work page 2012

  38. [38]

    Proceedings of the Second Workshop on Semantic Interpretation in an Actionable Context. 2012

    work page 2012

  39. [39]

    Learning to Interpret Natural Language Instructions

    Babe s -Vroman, Monica and MacGlashan, James and Gao, Ruoyuan and Winner, Kevin and Adjogah, Richard and desJardins, Marie and Littman, Michael and Muresan, Smaranda. Learning to Interpret Natural Language Instructions. Proceedings of the Second Workshop on Semantic Interpretation in an Actionable Context. 2012

    work page 2012

  40. [40]

    Toward Learning Perceptually Grounded Word Meanings from Unaligned Parallel Data

    Tellex, Stefanie and Thaker, Pratiksha and Joseph, Josh and Roy, Nicholas. Toward Learning Perceptually Grounded Word Meanings from Unaligned Parallel Data. Proceedings of the Second Workshop on Semantic Interpretation in an Actionable Context. 2012

    work page 2012

  41. [41]

    Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  42. [42]

    Measuring the Influence of Long Range Dependencies with Neural Network Language Models

    Le, Hai Son and Allauzen, Alexandre and Yvon, Fran c ois. Measuring the Influence of Long Range Dependencies with Neural Network Language Models. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  43. [43]

    Large, Pruned or Continuous Space Language Models on a GPU for Statistical Machine Translation

    Schwenk, Holger and Rousseau, Anthony and Attik, Mohammed. Large, Pruned or Continuous Space Language Models on a GPU for Statistical Machine Translation. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  44. [44]

    and Kingsbury, Brian and Ramabhadran, Bhuvana

    Arisoy, Ebru and Sainath, Tara N. and Kingsbury, Brian and Ramabhadran, Bhuvana. Deep Neural Network Language Models. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  45. [45]

    A Challenge Set for Advancing Language Modeling

    Zweig, Geoffrey and Burges, Chris J.C. A Challenge Set for Advancing Language Modeling. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  46. [46]

    Unsupervised Vocabulary Adaptation for Morph-based Language Models

    Mansikkaniemi, Andr \'e and Kurimo, Mikko. Unsupervised Vocabulary Adaptation for Morph-based Language Models. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  47. [47]

    Large-scale discriminative language model reranking for voice-search

    Jyothi, Preethi and Johnson, Leif and Chelba, Ciprian and Strope, Brian. Large-scale discriminative language model reranking for voice-search. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  48. [48]

    Revisiting the Case for Explicit Syntactic Information in Language Models

    Rastrow, Ariya and Khudanpur, Sanjeev and Dredze, Mark. Revisiting the Case for Explicit Syntactic Information in Language Models. Proceedings of the NAACL - HLT 2012 Workshop: Will We Ever Really Replace the N-gram Model? On the Future of Language Modeling for HLT. 2012

    work page 2012

  49. [49]

    Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  50. [50]

    and Dang, Hoa Trang and Nenkova, Ani

    Owczarzak, Karolina and Conroy, John M. and Dang, Hoa Trang and Nenkova, Ani. An Assessment of the Accuracy of Automatic Evaluation in Summarization. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  51. [51]

    Using the Omega Index for Evaluating Abstractive Community Detection

    Murray, Gabriel and Carenini, Giuseppe and Ng, Raymond. Using the Omega Index for Evaluating Abstractive Community Detection. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  52. [52]

    Machine Translation for Multilingual Summary Content Evaluation

    Steinberger, Josef and Turchi, Marco. Machine Translation for Multilingual Summary Content Evaluation. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  53. [53]

    Ecological Validity and the Evaluation of Speech Summarization Quality

    McCallum, Anthony and Munteanu, Cosmin and Penn, Gerald and Zhu, Xiaodan. Ecological Validity and the Evaluation of Speech Summarization Quality. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  54. [54]

    The Heterogeneity Principle in Evaluation Measures for Automatic Summarization

    Amig \'o , Enrique and Gonzalo, Julio and Verdejo, Felisa. The Heterogeneity Principle in Evaluation Measures for Automatic Summarization. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  55. [55]

    Discrepancy Between Automatic and Manual Evaluation of Summaries

    Mithun, Shamima and Kosseim, Leila and Perera, Prasad. Discrepancy Between Automatic and Manual Evaluation of Summaries. Proceedings of Workshop on Evaluation Metrics and System Comparison for Automatic Summarization. 2012

    work page 2012

  56. [56]

    Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  57. [57]

    Computational Analysis of Referring Expressions in Narratives of Picture Books

    Lee, Choonkyu and Muresan, Smaranda and Stromswold, Karin. Computational Analysis of Referring Expressions in Narratives of Picture Books. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  58. [58]

    A Computational Analysis of Style, Affect, and Imagery in Contemporary Poetry

    Kao, Justine and Jurafsky, Dan. A Computational Analysis of Style, Affect, and Imagery in Contemporary Poetry. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  59. [59]

    Towards a Literary Machine Translation: The Role of Referential Cohesion

    Voigt, Rob and Jurafsky, Dan. Towards a Literary Machine Translation: The Role of Referential Cohesion. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  60. [60]

    Unsupervised Stylistic Segmentation of Poetry with Change Curves and Extrinsic Features

    Brooke, Julian and Hammond, Adam and Hirst, Graeme. Unsupervised Stylistic Segmentation of Poetry with Change Curves and Extrinsic Features. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  61. [61]

    Aligning Bilingual Literary Works: a Pilot Study

    Yu, Qian and Max, Aur \'e lien and Yvon, Fran c ois. Aligning Bilingual Literary Works: a Pilot Study. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  62. [62]

    Function Words for C hinese Authorship Attribution

    Yu, Bei. Function Words for C hinese Authorship Attribution. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  63. [63]

    Mining wisdom

    S gaard, Anders. Mining wisdom. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  64. [64]

    Literary authorship attribution with phrase-structure fragments

    van Cranenburgh, Andreas. Literary authorship attribution with phrase-structure fragments. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  65. [65]

    Digitizing 18th-Century F rench Literature: Comparing transcription methods for a critical edition text

    Irvine, Ann and Marcellesi, Laure and Zomorodian, Afra. Digitizing 18th-Century F rench Literature: Comparing transcription methods for a critical edition text. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  66. [66]

    A Dictionary of Wisdom and Wit: Learning to Extract Quotable Phrases

    Bendersky, Michael and Smith, David. A Dictionary of Wisdom and Wit: Learning to Extract Quotable Phrases. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  67. [67]

    A Pilot P rop B ank Annotation for Q uranic A rabic

    Zaghouani, Wajdi and Hawwari, Abdelati and Diab, Mona. A Pilot P rop B ank Annotation for Q uranic A rabic. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  68. [68]

    Trend Analysis in Harper ' s Bazaar

    Kushkuley, Sophie. Trend Analysis in Harper ' s Bazaar. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  69. [69]

    Social Network Analysis of Alice in Wonderland

    Agarwal, Apoorv and Corvalan, Augusto and Jensen, Jacob and Rambow, Owen. Social Network Analysis of Alice in Wonderland. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  70. [70]

    Towards a computational approach to literary text analysis

    Roque, Antonio. Towards a computational approach to literary text analysis. Proceedings of the NAACL - HLT 2012 Workshop on Computational Linguistics for Literature. 2012

    work page 2012

  71. [71]

    B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  72. [72]

    Graph-based alignment of narratives for automated neurological assessment

    Prud ' hommeaux, Emily and Roark, Brian. Graph-based alignment of narratives for automated neurological assessment. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  73. [73]

    Bootstrapping Biomedical Ontologies for Scientific Text using NELL

    Movshovitz-Attias, Dana and Cohen, William W. Bootstrapping Biomedical Ontologies for Scientific Text using NELL. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  74. [74]

    Semantic distance and terminology structuring methods for the detection of semantically close terms

    Dupuch, Marie and Dupuch, La. Semantic distance and terminology structuring methods for the detection of semantically close terms. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  75. [75]

    Temporal Classification of Medical Events

    Raghavan, Preethi and Fosler-Lussier, Eric and Lai, Albert. Temporal Classification of Medical Events. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  76. [76]

    Analyzing Patient Records to Establish If and When a Patient Suffered from a Medical Condition

    Cogley, James and Stokes, Nicola and Carthy, Joe and Dunnion, John. Analyzing Patient Records to Establish If and When a Patient Suffered from a Medical Condition. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  77. [77]

    Alignment- HMM -based Extraction of Abbreviations from Biomedical Text

    Movshovitz-Attias, Dana and Cohen, William W. Alignment- HMM -based Extraction of Abbreviations from Biomedical Text. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  78. [78]

    Medical diagnosis lost in translation -- Analysis of uncertainty and negation expressions in E nglish and S wedish clinical texts

    Mowery, Danielle L and Velupillai, Sumithra and Chapman, Wendy W. Medical diagnosis lost in translation -- Analysis of uncertainty and negation expressions in E nglish and S wedish clinical texts. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  79. [79]

    A Hybrid Stepwise Approach for De-identifying Person Names in Clinical Documents

    Ferr \'a ndez, Oscar and South, Brett and Shen, Shuying and Meystre, St \'e phane. A Hybrid Stepwise Approach for De-identifying Person Names in Clinical Documents. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

  80. [80]

    Active Learning for Coreference Resolution

    Miller, Timothy and Dligach, Dmitriy and Savova, Guergana. Active Learning for Coreference Resolution. B io NLP : Proceedings of the 2012 Workshop on Biomedical Natural Language Processing. 2012

    work page 2012

Showing first 80 references.