Representing text as abstract images enables image classifiers to also simultaneously classify text

Stephen M. Petrie; T'Mir D. Julius

REVIEW 3 major objections 4 minor 32 references

Representing text as abstract images enables image classifiers to also simultaneously classify text

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A text-to-pixels map lets image classifiers match names at 99.09% F1

desk verdict The comparison-map encoding is a genuinely new idea, but the reported F1 is not credible as a state-of-the-art claim because the test set was used to tune thresholds and the record-level split leaks inventor identity across train and test. read the letter →

arxiv 1908.07846 v3 pith:BYNSHACT submitted 2019-08-19 cs.CL cs.AIcs.CVcs.LG

Stephen M. Petrie , T'Mir D. Julius This is my paper · ORCID

classification cs.CLcs.AIcs.CVcs.LG

keywords inventornamedisambiguationtext-to-imagerepresentationcomparison-mapimagesconvolutionalneuralnetworkspairwisetextclassificationpatentdatarecordlinkageentity

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that text can be converted into a two-dimensional RGB image — a “comparison-map” — so that standard image classifiers can perform text classification. Applying the idea to inventor-name disambiguation in US patent records, it renders each pair of inventor records as a single stacked image and trains the AlexNet convolutional network to label the pair as same inventor or different inventors. The network reaches F1 99.09%, precision 99.41%, and recall 98.76% on the combined labeled benchmark, marginally above the previous state of the art. The appeal is that the classifier learns its own features from the images, rather than from hand-chosen string-similarity measures, and the same representation is meant to carry over to other text-matching tasks such as author disambiguation.

What carries the argument

The comparison-map image carries the argument. It is built by producing a string-map for each text field — a 2D grid of letter positions with colored pixels along each letter and along lines connecting the letters of each bigram — then stacking two records' maps as the red and green channels of one RGB tensor, with the first bigram of each record sharing the blue channel. This turns string similarity into spatial overlap: matching text produces yellow, mismatching text produces separate red and green regions. Because the result is a single image, any image classification network can be used; the paper uses a minimally modified AlexNet whose input layer, first-layer kernel size and stride, and output layer are adjusted for 31-by-31-by-3 inputs and two classes.

What would settle it

Take the IS and E&S labeled data, group records by inventor ID, and split so that no inventor appears in both training and test data; then retrain and re-evaluate, tuning the probability and linkage thresholds on a separate validation split. If the F1 falls materially below 99.09%, the headline number is inflated by within-inventor leakage and test-set tuning.

Watch

Extended reading notes

Core claim

The central discovery is that a pairwise text comparison can be rendered as a single RGB image, and an off-the-shelf image classifier then performs the text comparison without any text-specific architecture. Each field of an inventor record — first name, middle name, last name, city, IPC codes, co-inventors, assignees — is drawn onto a fixed 2D character grid as colored letter shapes with connecting lines; the two records being compared are placed in the red and green channels of the same image, with the leading two-letter bigram highlighted in blue. Similar records produce large yellow overlap zones, while dissimilar records stay mostly red and green, and the convolutional network picks up these pixel-level cues. On the combined Israeli-inventor and engineer/scientist labeled data, the method reaches F1 99.09%, precision 99.41%, and recall 98.76%, and the authors report that randomizing the character order, the pixel layout, or both changes the F1 only within 98.99% to 99.09%. The discovery, if sound, is that visual pattern recognition can substitute for hand-engineered string similarity in text-pair decisions.

Load-bearing premise

The reported F1 assumes that the random record-level split keeps every inventor's records entirely in either the training set or the test set; if records from the same inventor appear in both, the score partly measures memorization of known individuals rather than generalization to unseen inventors, and the tuning choices are also made on the same test set.

Editorial extensions

If this is right

Image classification models that already exist can be applied to text-pair classification tasks with minor hyperparameter changes, so gains in image modeling become available for NLP-style matching problems.
The approach avoids hand-selected string-similarity features; the network learns its own discriminative cues directly from the rendered maps.
If the central claim holds, the same comparison-map construction should transfer to other pairwise text tasks, such as academic author disambiguation and record linkage, as the paper argues.
The small performance differences across randomized string-map layouts indicate that the method does not depend on a carefully tuned glyph arrangement.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural holdout experiment the paper does not report is splitting the labeled data by inventor rather than by record; comparing F1 under that split would show how much of the 99.09% is generalization to unseen people versus recognition of individuals already seen in training.
Because the representation is already a tensor, the same stacking trick could be extended to mixed-modal records — for instance, a product image in one channel and its textual description in another — though the paper only demonstrates text-on-text comparisons.
The layout-robustness result suggests the method may transfer to alphabets or scripts very different from English, but the paper tests only Latin-script names and alphanumeric codes, leaving cross-script behavior open.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The thing to know: the paper has a genuinely novel encoding for pairwise text comparison. It turns two text records into RGB images by stacking record-maps in the red and green channels, with line-drawn letter links and a heuristic letter layout. That is not a trivial restatement of character-level CNNs or siamese networks, and I don't know of prior work doing this for record linkage. Credit where due: the authors test several random letter layouts and get nearly identical F1 scores (98.99–99.09%), which suggests the representation itself is capturing useful structure rather than depending on a lucky layout. The paper is also clearly written and transparent about its appendices, blocking, and clustering steps.

The soft spots are real and appear in the paper's own description. In Section V-A they state that the thresholds p_bar and l_bar were chosen by maximizing F1 on the labelled test data. That makes the reported 99.09% an in-sample estimate, not an unbiased evaluation. The effect is modest because the F1 spread across layouts is tiny, but the claimed edge over Kim et al. (99.09 vs 99.01) is within the noise of that selection. Bigger problem: the 80/20 split is at the record level, not the inventor level. The IS and E&S datasets contain multiple patent records per inventor, so the same inventor can appear in both training and test. The comparison-map encodes name, city, co-inventors, assignees, and IPC codes, so the CNN can memorize specific inventor patterns and then recognize their remaining records in the test set. The reported F1 therefore measures recognition of already-seen people, not generalization to unseen inventors—which is what bulk disambiguation actually needs. This issue is load-bearing for the quantitative claim, not a minor footnote.

That said, the core idea still deserves attention. The encoding is a clean way to apply off-the-shelf image classifiers to text comparison, and the robustness to random layouts is a point in its favor. The evaluation problems are fixable: split by inventor, select thresholds on a validation set, and report the spread over runs rather than the best. Until then, I would treat the F1 numbers skeptically but the representation as a useful contribution. Who is this for? People working on entity disambiguation or record linkage who want an alternative to hand-engineered string features. They will get value from the encoding, but should not quote the SOTA comparison. I would send it to peer review because the idea is worth publishing after major revision on the evaluation.

Referee Report

3 major / 4 minor

Summary. The paper proposes a method for converting text-based records into abstract 2D RGB 'comparison-map' images, allowing a standard image classifier (AlexNet) to perform pairwise match/non-match classification. The method is applied to inventor name disambiguation in USPTO patent data. The authors report an F1 score of 99.09% on the combined IS and E&S labeled datasets, marginally outperforming prior work by Kim et al. (F1 = 99.01%). The paper also tests several alternative string-map layouts and reports F1 scores between 98.99% and 99.09%, suggesting the representation is robust to layout choices.

Significance. If the evaluation were sound, the contribution would be significant: the text-to-image representation is a creative way to reuse mature image-classification architectures for text-based pairwise comparison tasks, enabling end-to-end feature learning without hand-crafted string similarity features. The robustness analysis across randomized string-map layouts is a useful strength, as is the explicit reporting of precision and recall components. However, the central quantitative claim—that the method outperforms the previous state of the art—is compromised by two evaluation-protocol issues: the clustering thresholds are tuned on the test set, and the train/test split is performed at the record level without grouping by inventor, allowing potential identity leakage. These issues must be resolved before the reported F1 can be considered a trustworthy estimate of generalization performance.

major comments (3)

[V-A and V-C] This is not a minor caveat: the entire quantitative claim, including the marginal advantage over Kim et al. (99.09% vs. 99.01%), depends on parameters that were chosen by looking at the test labels. The magnitude of the resulting bias is unknown but could easily exceed the 0.08 percentage-point difference that the paper highlights.
[V-A] The paper does not report how many inventors appear in both partitions, nor does it report the number of records per inventor. Without such information, the extent of leakage cannot be assessed. This issue is load-bearing because the primary claim is about the method's ability to disambiguate previously unseen inventor records.
[Appendix B] This is a secondary consequence of the split issue, but it is worth reporting separately because the authors explicitly use the ceiling to contextualize their recall values.

minor comments (4)

[Table III] The footnote for the 'Both' rows states that the IS and E&S results for prior work were combined by averaging, while the 'Ours' results appear to come from a single run on a combined dataset. This is not an apples-to-apples comparison; the paper should state explicitly whether the authors' Both result is from a model trained and evaluated on the union of the two datasets or from averaging separate runs, and should ensure consistent aggregation for all methods.
[V-D] The paper reports that for each alternative string-map, the algorithm was run multiple times with different threshold settings and only the run with the highest F1 is shown. This is another instance of test-set selection, and although it is partly covered by the major comment on threshold selection, the presentation currently reads as if the alternative string-map comparison is a fair robustness check; it would be clearer to report the mean and variance over threshold choices or to use a fixed, pre-specified threshold selection rule.
[IV-B] The statement that 'we do not use any of those transformations' in DIGITS is ambiguous, because the default settings of the implementation may still apply some preprocessing unless explicitly disabled. Please clarify whether all three default transformations (mean subtraction, random mirroring, random cropping) were fully disabled in the experiments.
[Abstract and I] The phrase 'simultaneously classify text' in the abstract is likely intended to mean 'classify text using an image classifier,' but it reads as if the network classifies text and images at the same time. Please rephrase to avoid ambiguity, as the actual contribution is about applying an image classifier to text-derived images, not multimodal classification.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported F1 is a test-set-maximized statistic: clustering thresholds are tuned on the same labelled test data used to compute the headline performance.

fitted input called prediction [Section V-A, paragraph on threshold selection before Section V-B]
"Prior to processing the bulk data, we experimented with multiple different values for the pairwise comparison probability threshold ( ¯p) and linking proportion threshold ( ¯l), based on evaluating the trained neural network on the labelled test data. Different ¯p and ¯l values produce different trade-offs between precision and recall, and we use values that produce an optimal trade-off (highest F1 score)."

The paper selects the clustering thresholds p_bar and l_bar by maximizing F1 on the labelled test data, then reports F1 computed on that same test data as the algorithm's performance (e.g., 99.09% in Table III). The reported F1 is therefore the maximum of a grid search over thresholds evaluated on the test set, not an unbiased hold-out estimate. The performance 'prediction' is partly forced by construction: the threshold choice is fitted to the exact metric (F1) on the exact data used for evaluation.

full rationale

The paper's core qualitative contribution—converting text into abstract image representations so that image classifiers can perform text classification—is not itself circular: the representation is defined independently of the match/non-match labels, the network is trained on labelled comparison-maps, and robustness checks across random string-map layouts provide independent evidence that performance does not depend on the specific heuristic layout. There are no load-bearing self-citations: the cited methods (AlexNet, Kim et al., etc.) are external, and no uniqueness or ansatz is imported from the authors' prior work. However, the central quantitative claim is partially circular under the fitted-input-called-prediction pattern. The paper explicitly tunes p_bar and l_bar by maximizing F1 on the same labelled test data that is then used to report F1, so the headline numbers (F1 99.09%, precision 99.41%, recall 98.76%) are maximized over thresholds on the evaluation set. This is compounded by the record-level random split, which can place records of the same inventor in both training and test sets, allowing the CNN to memorize inventor-specific name/address/assignee/co-inventor patterns. These are evaluation-validity problems rather than derivation-identity circularity of the representation method itself, but they directly undermine the paper's main performance prediction. Score 6 reflects partial circularity in the central quantitative claim, while acknowledging that the representation idea retains independent content.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central method is a new data representation rather than a theory with axioms. The main free parameters are the two thresholds used to convert probabilities into clusters, both of which were tuned on the test set, and the block-size threshold. The paper assumes the labelled data is correct, the blocking retains true pairs, and the encoding preserves sufficient information for learning.

free parameters (3)

p_bar (pairwise match probability threshold) = 0.03 (for best F1 run)
Used to binarize pairwise match probabilities (Equation 6); selected by grid search to maximize F1 on the labelled test data (Section V-A).
l_bar (group linking proportion threshold) = 0.05 (for best F1 run)
Used in clustering to decide whether a record is kept in a group; selected jointly with p_bar on the test data (Appendix C).
n_bar (block size threshold) = 100
Controls blocking granularity (Appendix B); chosen after initial testing as a compromise between efficiency and recall, but not central to the final accuracy.

assumptions (4)

domain assumption The IS and E&S labelled datasets provide accurate ground truth for inventor identity.
All training and evaluation depends on the correctness of these labels; no independent verification is provided.
domain assumption Blocking by first letters of last name retains nearly all true matches within blocks.
The algorithm only compares records within blocks; the paper estimates a maximum recall limit in Section V-B (99.47% to 99.98%), so this is an acknowledged approximation.
domain assumption The string-map encoding preserves enough textual information for the CNN to learn match/non-match discrimination.
This is the core modeling assumption of the method; it is validated only empirically through the reported results and the random-layout robustness tests in Section V-D.
domain assumption The 80/20 random split yields test records that are independent of training records.
Section V-A describes a random split without grouping by inventor; if inventors overlap splits, this assumption is violated.

invented entities (1)

Comparison-map image (string-map and record-map)
purpose: Encodes pairs of text records as stacked RGB images so that a CNN can classify matches.
The representation is introduced by this paper; its utility is demonstrated only in the paper's own experiments, with no external falsifiable handle.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Representing text as abstract images enables image classifiers to also simultaneously classify text." pith.science (2026). https://pith.science/paper/BYNSHACT

@misc{pith2026190807846,
  author       = {Pith},
  title        = {Pith review of: Representing text as abstract images enables image classifiers to also simultaneously classify text},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BYNSHACT}},
  note         = {Machine review of arXiv:1908.07846}
}

read the original abstract

We introduce a novel method for converting text data into abstract image representations, which allows image-based processing techniques (e.g. image classification networks) to be applied to text-based comparison problems. We apply the technique to entity disambiguation of inventor names in US patents. The method involves converting text from each pairwise comparison between two inventor name records into a 2D RGB (stacked) image representation. We then train an image classification neural network to discriminate between such pairwise comparison images, and use the trained network to label each pair of records as either matched (same inventor) or non-matched (different inventors), obtaining highly accurate results. Our new text-to-image representation method could also be used more broadly for other NLP comparison problems, such as disambiguation of academic publications, or for problems that require simultaneous classification of both text and image datasets.

Figures

Figures reproduced from arXiv: 1908.07846 by the authors.

**Figure 1.** Constructing a string-map image. The first four images show the sub-maps that are summed to construct the final string-map image (right-most image), for the example word “JEN”. 4) Train neural network: use 2D comparison-map images generated from manually labelled data to train a neural network to classify whether a given pairwise record comparison is a match (same inventor) or non-match (different inventors). 5) Cl… view at source ↗

**Figure 2.** Comparison of two strings. To compare the names “JEN” and “LIN”, we add the string-map for “JEN” (left image) to the string-map for “LIN” (middle image) to generate the final comparison image (right image). “string-map” is shown in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Larger string-map for assignees and co-inventors, and IPC-map. The larger string-map used to convert a given list of assignees or co-inventors into an abstract image representation (left), and the IPC-map used to convert a given list of IPC classes into an abstract image representation (right) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Record-map layout. Shows the positioning of each stringmap within a given record-map. colour to each pixel in these larger string-maps, again to reduce the possibility of saturation. For international patent classification (IPC) codes, which contain numbers as well as…

**Figure 5.** Figure 5: Comparison-map examples. Two examples of comparison-map images. The left comparison-map image was generated using two matched records (Table I, rows 1 and 2), and the right image from two non-matched records (Table I, rows 1 and 3). parison problems (e.g. data linkage,…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

[1]

Collective knowledge, proliﬁc inventors and the value of inventions: An empirical study of French, German and British patents in the US, 1975- 1999,

C. Gay, W. Latham, and C. Le Bas, “Collective knowledge, proliﬁc inventors and the value of inventions: An empirical study of French, German and British patents in the US, 1975- 1999,” Economics of Innovation and New Technology , vol. 17, no. 1-2, pp. 5–22, 2008

work page 1975
[2]

Does mobility increase the productivity of inven- tors?

K. Hoisl, “Does mobility increase the productivity of inven- tors?” Journal of Technology Transfer, vol. 34, no. 2, pp. 212– 225, 2009

work page 2009
[3]

Disambiguation and co-authorship networks of the U.S. patent inventor database (1975-2010),

G. C. Li, R. Lai, A. D’Amour, D. M. Doolin, Y . Sun, V . I. Torvik, A. Z. Yu, and F. Lee, “Disambiguation and co-authorship networks of the U.S. patent inventor database (1975-2010),” Research Policy, vol. 43, no. 6, pp. 941–955,

work page 1975
[4]

Seeing the non- stars: (Some) sources of bias in past disambiguation approaches and a new public tool leveraging labeled records,

S. L. Ventura, R. Nugent, and E. R. H. Fuchs, “Seeing the non- stars: (Some) sources of bias in past disambiguation approaches and a new public tool leveraging labeled records,” Research Policy, vol. 44, no. 9, pp. 1672–1701, 2015. [Online]. Available: http://linkinghub.elsevier.com/retrieve/pii/S0048733314002406

work page 2015
[5]

Random Forest DBSCAN for USPTO Inventor Name Disambiguation

K. Kim, M. Khabsa, and C. L. Giles, “Random Forest DBSCAN for USPTO Inventor Name Disambiguation,” arXiv:1602.01792, vol. [cs.IR], 2016. [Online]. Available: http://arxiv.org/abs/1602.01792

work page Pith review arXiv 2016
[6]

The ”Names Game

M. Trajtenberg, G. Shiff, and R. Melamed, “The ”Names Game”: Harnessing Inventors’ Patent Data for Economic Re- search,”National Bureau of Economic Research Working Paper, vol. 12479, 2006

work page 2006
[7]

Singling Out Individual Inventors from Patent Data,

E. Migu ´elez and I. G ´omez-Migu´elez, “Singling Out Individual Inventors from Patent Data,” Research Institute of Applied Economics Working Paper, 2011

work page 2011
[8]

ImageNet Clas- siﬁcation with Deep Convolutional Neural Networks,

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Clas- siﬁcation with Deep Convolutional Neural Networks,”Advances in Neural Information Processing Systems , vol. 25, pp. 1097– 1105, 2012

work page 2012

Show all 32 references

[9]

How to play the

J. Raffo and S. Lhuillery, “How to play the ”Names Game”: Patent retrieval comparing different heuristics,”Research Policy, vol. 38, no. 10, pp. 1617–1627, 2009

2009
[10]

How to kill inven- tors: testing the Massacrator algorithm for inventor disambigua- tion,

M. Pezzoni, F. Lissoni, and G. Tarasconi, “How to kill inven- tors: testing the Massacrator algorithm for inventor disambigua- tion,” Scientometrics, vol. 101, no. 1, pp. 477–504, 2014

2014
[11]

Binary codes capable of correcting deletions, insertions, and reversals,

V . I. Levenshtein, “Binary codes capable of correcting deletions, insertions, and reversals,”Soviet Physics Doklady, vol. 10, no. 8, pp. 707–710, 1966

1966
[12]

Advances in record-linkage methodology as applied to matching the 1985 census of Tampa, Florida,

M. a. Jaro, “Advances in record-linkage methodology as applied to matching the 1985 census of Tampa, Florida,” Journal of the American Statistical Association, vol. 84, no. 406, pp. 414–420, 1989

1985
[13]

String Comparator Metrics and Enhanced Decision Rules in the Fellegi-Sunter Model of Record Linkage,

W. E. Winkler, “String Comparator Metrics and Enhanced Decision Rules in the Fellegi-Sunter Model of Record Linkage,” Proceedings of the American Statistical Association Section on Survey Research Methods , 1990. [Online]. Available: http://scholar.google.com/scholar?hl=en&btn...

1990
[14]

Disambiguation of patent inventors and assignees using high-resolution geolo- cation data,

G. Morrison, M. Riccaboni, and F. Pammolli, “Disambiguation of patent inventors and assignees using high-resolution geolo- cation data,” Scientiﬁc Data, vol. 4, pp. 1–21, 2017

2017
[15]

Character-level Convolutional Networks for Text Classiﬁcation,

X. Zhang, J. Zhao, and Y . LeCun, “Character-level Convolutional Networks for Text Classiﬁcation,” Advances in Neural Information Processing Systems , pp. 1–9, 2015. [Online]. Available: http://arxiv.org/abs/1502.01710

2015 arXiv
[16]

Siamese Neural Networks for One-Shot Image Recognition,

G. Koch, R. Zemel, and R. Salakhutdinov, “Siamese Neural Networks for One-Shot Image Recognition,” Proceedings of the 32nd International Conference on Machine Learning , vol. 37, 2015

2015
[17]

Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches,

J. Zbontar and Y . LeCun, “Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches,” Journal of Machine Learning Research, vol. 17, pp. 1–32, 2016

2016
[18]

Learning to Compare Image Patches via Convolutional Neural Networks,

S. Zagoruyko and N. Komodakis, “Learning to Compare Image Patches via Convolutional Neural Networks,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2015

2015
[19]

ABCNN: Attention-Based Convolutional Neural Network for Modeling Sentence Pairs,

W. Yin, H. Sch ¨utze, B. Xiang, and B. Zhou, “ABCNN: Attention-Based Convolutional Neural Network for Modeling Sentence Pairs,” Transactions of the Association for Computational Linguistics , vol. 4, pp. 259–272, 2016. [Online]. Available: http://arxiv.org/abs/1512.05193

2016 arXiv
[20]

Signature Veriﬁca- tion Using a Siamese Time Delay Neural Network,

J. Bromley, J. W. Bentz, L. Bottou, I. Guyon, Y . Lecun, C. Moore, E. S ¨ackinger, and R. Shah, “Signature Veriﬁca- tion Using a Siamese Time Delay Neural Network,” Interna- tional Journal of Pattern Recognition and Artiﬁcial Intelligence, vol. 07, no. 04, pp. 669–688, 1993

1993
[21]

Discriminative deep metric learning for face veriﬁcation in the wild,

J. Hu, J. Lu, and Y . P. Tan, “Discriminative deep metric learning for face veriﬁcation in the wild,” Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 1875–1882, 2014

2014
[22]

The NBER Patent Citation Data File: Lessons, Insights and Methodological Tools,

B. H. Hall, A. B. Jaffe, and M. Trajtenberg, “The NBER Patent Citation Data File: Lessons, Insights and Methodological Tools,”National Bureau of Economic Research Working Paper , vol. 8498, 2001. [Online]. Available: http: //www.nber.org/papers/w8498.pdf?new window=1

2001
[23]

Engineer/scientist careers: Patents, online proﬁles, and misclassiﬁcation bias,

C. Ge, K. Huang, and I. P. L. Png, “Engineer/scientist careers: Patents, online proﬁles, and misclassiﬁcation bias,” Strategic Management Journal, vol. 37, pp. 232–253, 2016

2016
[24]

Caffe: Convolutional Architecture for Fast Feature Embedding,

Y . Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarrama, and T. Darrell, “Caffe: Convolutional Architecture for Fast Feature Embedding,” Proceedings of the 22nd ACM international conference on Multimedia , pp. 675–678, 2014. [Online]. Available: http...

2014 arXiv
[25]

A Mixture Record Linkage Approach for US Patent Inven- tor Disambiguation,

G.-C. Yang, C. Liang, Z. Jing, D.-R. Wang, and H.-C. Zhang, “A Mixture Record Linkage Approach for US Patent Inven- tor Disambiguation,” Advanced Multimedia and Ubiquitous Engineering. FutureTech 2017, MUE 2017. Lecture Notes in Electrical Engineering, vol. 448, pp. 331–338, 2...

2017
[27]

the number of asserted matches to other records), highest ﬁrst

Order all patent-inventor name records by the number of links they have to other records (i.e. the number of asserted matches to other records), highest ﬁrst
[28]

Assign a UID to each isolated (non-matched) patent-inventor name
[29]

That is, for a given record, the corresponding inventor group initially comprises just the record itself and all records it is linked (matched) to

Assign records to inventor groups. That is, for a given record, the corresponding inventor group initially comprises just the record itself and all records it is linked (matched) to. Each of these linked records (nodes) are kept in the current inventor group only if the number...
[30]

Repeat Step 2, because some records may have become isolated (non-matched) following Step 3
[31]

Combine inventor groups together if the number of links they share is greater than a speciﬁed threshold. In particular, for an inventor group with nself records (nodes), we combine it with any other group with nother nodes if the number of links to that other group ( l) satisﬁ...
[32]

For each resulting inventor group, assign an identi- cal UID to all patent-inventor name records in the group. D. Random string-map layouts Here we show the random string layouts analysed in Section V-D. Figure A1 shows the string-maps we use for runs where characters are posi...
[2014]

Available: http://dx.doi.org/10.1016/j.respol

[Online]. Available: http://dx.doi.org/10.1016/j.respol. 2014.01.012

2014 doi

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Collective knowledge, proliﬁc inventors and the value of inventions: An empirical study of French, German and British patents in the US, 1975- 1999,

C. Gay, W. Latham, and C. Le Bas, “Collective knowledge, proliﬁc inventors and the value of inventions: An empirical study of French, German and British patents in the US, 1975- 1999,” Economics of Innovation and New Technology , vol. 17, no. 1-2, pp. 5–22, 2008

work page 1975

[2] [2]

Does mobility increase the productivity of inven- tors?

K. Hoisl, “Does mobility increase the productivity of inven- tors?” Journal of Technology Transfer, vol. 34, no. 2, pp. 212– 225, 2009

work page 2009

[3] [3]

Disambiguation and co-authorship networks of the U.S. patent inventor database (1975-2010),

G. C. Li, R. Lai, A. D’Amour, D. M. Doolin, Y . Sun, V . I. Torvik, A. Z. Yu, and F. Lee, “Disambiguation and co-authorship networks of the U.S. patent inventor database (1975-2010),” Research Policy, vol. 43, no. 6, pp. 941–955,

work page 1975

[4] [4]

Seeing the non- stars: (Some) sources of bias in past disambiguation approaches and a new public tool leveraging labeled records,

S. L. Ventura, R. Nugent, and E. R. H. Fuchs, “Seeing the non- stars: (Some) sources of bias in past disambiguation approaches and a new public tool leveraging labeled records,” Research Policy, vol. 44, no. 9, pp. 1672–1701, 2015. [Online]. Available: http://linkinghub.elsevier.com/retrieve/pii/S0048733314002406

work page 2015

[5] [5]

Random Forest DBSCAN for USPTO Inventor Name Disambiguation

K. Kim, M. Khabsa, and C. L. Giles, “Random Forest DBSCAN for USPTO Inventor Name Disambiguation,” arXiv:1602.01792, vol. [cs.IR], 2016. [Online]. Available: http://arxiv.org/abs/1602.01792

work page Pith review arXiv 2016

[6] [6]

The ”Names Game

M. Trajtenberg, G. Shiff, and R. Melamed, “The ”Names Game”: Harnessing Inventors’ Patent Data for Economic Re- search,”National Bureau of Economic Research Working Paper, vol. 12479, 2006

work page 2006

[7] [7]

Singling Out Individual Inventors from Patent Data,

E. Migu ´elez and I. G ´omez-Migu´elez, “Singling Out Individual Inventors from Patent Data,” Research Institute of Applied Economics Working Paper, 2011

work page 2011

[8] [8]

ImageNet Clas- siﬁcation with Deep Convolutional Neural Networks,

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Clas- siﬁcation with Deep Convolutional Neural Networks,”Advances in Neural Information Processing Systems , vol. 25, pp. 1097– 1105, 2012

work page 2012