Pith. sign in

REVIEW 3 major objections 4 minor 91 references

FinTextSim: Enhancing Financial Text Analysis with BERTopic

T0 review · 3 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read BERTopic, a neural topic model, produces clear, distinct economic topic clusters from Item 7 and Item 7A of S&P 500 10-K filings only when paired with FinTextSim, a sentence-transformer fine-tuned on financial text; FinTextSim raises…

desk verdict A sensible domain-tuned embedding idea with real dataset work, but the FinTextSim-vs-AM comparison is confounded by base architecture and the headline metric is circular; worth a serious referee, not desk rejection, but it needs major revision. read the letter →

arxiv 2504.15683 v1 pith:BHIXCSVP submitted 2025-04-22 cs.CL cs.LGecon.GNq-fin.ECq-fin.GN

Simon Jehnen , Joaquín Ordieres-Meré , Javier Villalba-Díez This is my paper · ORCID
classification cs.CLcs.LGecon.GNq-fin.ECq-fin.GN
keywords FinTextSimBERTopictopicmodeling10-KfilingsfinancialanalysissentencetransformersMD&Atopic-precision
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 claims that domain-specific fine-tuning of sentence transformers is required for reliable topic modeling of financial disclosures. It introduces FinTextSim, a sentence-transformer fine-tuned on 180,435 keyword-labeled sentences from 10-K Item 7/Item 7A and other annual and sustainability reports, and reports that BERTopic only forms clear, distinct economic topic clusters when fed FinTextSim's embeddings rather than those of the widely used all-MiniLM-L6-v2. On the authors' 14-topic evaluation, FinTextSim achieves perfect topic-precision of 1.0 while the baseline misses up to ten of the fourteen topics, and it improves intratopic similarity by 81% while reducing intertopic similarity by 100%. The authors also report that raw NPMI coherence favors the baseline, and argue that coherence measures mislead for financial text because domain terms rarely co-occur in small windows. A sympathetic reader would take away that embedding choice drives topic quality more than any other BERTopic component.

What carries the argument

The load-bearing object is FinTextSim, a sentence-transformer built on ModernBERT with mean pooling and normalization, trained with adaptive circle loss, a curriculum variant of circle loss that shifts margin and scale during training. Its training data are 180,435 sentences labeled by a 14-topic financial keyword list derived from prior accounting and economics literature, augmented with Energy, ESG, and COVID-19 topics. The same keyword list labels the training data, seeds BERTopic's topic-word extraction with a multiplier of 50, and defines topic-precision (Equation 1), the weighting factor multiplied into NPMI coherence and similarity scores and divided out of intertopic similarity. The BERTopic pipeline uses UMAP to ten components, HDBSCAN with minimum cluster size 1,250, CountVectorizer with the Loughran-McDonald stopword list, and c-tfidf weighting. The paper's comparisons hold every BERTopic component fixed except the embedding model.

What would settle it

Re-run the BERTopic comparison with topic-precision computed from a held-out keyword list that played no role in training or topic seeding, or have independent financial analysts label the inferred topics; alternatively, feed FinTextSim embeddings of an out-of-domain financial corpus such as analyst reports or earnings-call transcripts into BERTopic and check whether the clusters stay as clean as on 10-K text. If the gap over the baseline narrows or vanishes, the reported advantage comes from the shared keyword scheme rather than the embeddings.

Watch

Extended reading notes

Core claim

The central claim is that the quality of BERTopic's topics on Item 7 and Item 7A of S&P 500 10-K filings is determined by the embedding model, and that general-purpose embeddings fail at this task while FinTextSim succeeds. With FinTextSim, all fourteen economic topics in the authors' scheme are recovered with topic-precision 1.0 on both sentence and refined-sentence input; intratopic similarity reaches 0.9972 versus 0.5498 for all-MiniLM-L6-v2, and intertopic similarity drops to 0.0002 versus 0.4647. The paper also reports that the baseline model leaves 226,605 of 687,959 sentences as BERTopic outliers, versus 184,470 for FinTextSim. Notably, raw NPMI coherence is higher for the baseline in every configuration; the authors attribute this to the many outlier sentences simplifying topic compression and to financial vocabulary not co-occurring within sliding windows, and conclude that coherence without domain-weighting is insufficient for evaluating financial topics.

Load-bearing premise

The load-bearing premise is that topic-precision, computed from the same 14-topic keyword list used to label FinTextSim's training sentences, seed BERTopic's topic words, and score the results, is a valid measure of financial topic quality; if that metric does not reflect real topic quality, the perfect 1.0 score collapses because the model is being graded on labels it was built and guided to reproduce.

Editorial extensions

If this is right

  • FinTextSim's reported gains imply that embedding choice, not clustering or weighting, is the main lever on topic quality for 10-K MD&A text.
  • Downstream tasks that depend on topic separation, such as aspect-based sentiment analysis of filings, business valuation, and stock-price prediction, would inherit cleaner inputs if the reported separation holds.
  • Raw NPMI coherence would no longer be treated as sufficient evidence of topic quality in finance; precision-weighted metrics would be needed to avoid misleading rankings.
  • If the approach generalizes beyond Item 7/7A, the same fine-tuning recipe could be applied to other 10-K sections and to other regulated disclosure documents.

Reading between the lines

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

  • The near-total collapse of intertopic similarity from 0.4647 to 0.0002 suggests FinTextSim's embedding space may be collapsed onto the 14 training labels; a testable consequence is that topics outside the keyword scheme would be poorly separated, which a reader could check with a held-out topic.
  • Because the same keyword list trains the model and defines the evaluation, the perfect 1.0 topic-precision should be read as an upper bound; an independent annotation study would be the natural way to confirm the paper's main claim.
  • The reported 19% reduction in BERTopic outliers suggests the fine-tuned geometry is denser around the labeled topics; an interesting extension is measuring whether that density transfers to unlabeled financial genres like conference-call transcripts.
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.

Referee Report

3 major / 4 minor

Summary. The paper introduces FinTextSim, a sentence-transformer model fine-tuned on financial text (based on ModernBERT), and evaluates BERTopic topic modeling on Item 7 and Item 7A of 10-K filings from S&P 500 companies. The authors compare FinTextSim against all-MiniLM-L6-v2 (AM), reporting that FinTextSim increases intratopic similarity by 81% and reduces intertopic similarity by 100%, and that BERTopic forms clear economic topic clusters only when using FinTextSim embeddings. The evaluation relies on a topic-precision metric (Eq. 1) computed from the same 14-topic keyword list used to create FinTextSim's training labels, to select refined sentences for BERTopic, and to seed BERTopic's c-tfidf with a multiplier of 50.

Significance. If the claims were sound, the paper would make a useful applied contribution by showing that domain-specific fine-tuning of sentence transformers materially improves topic modeling on financial disclosures. The work addresses a relevant gap in applying neural topic models to MD&A text, and it includes qualitative examples and a detailed pipeline. However, the central quantitative evidence is undermined by two load-bearing problems: the main evaluation metric is circular with respect to the training and seeding procedure, and the model comparison conflates architectural differences with the effect of fine-tuning. These issues affect the paper's main conclusions, not just peripheral details.

major comments (3)
  1. [§3.5.3, Eq. (1); Tables 3 and 4] The topic-precision metric is defined from the same 14-topic keyword list that is used to create FinTextSim's training labels (§3.3), to select refined sentences (§3.4), and to seed BERTopic's c-tfidf with a multiplier of 50 (§3.4). Consequently, the perfect topic-precision score of 1.0 in Table 3 is largely by construction: the model is rewarded for reproducing the authors' own taxonomy, not for discovering independently meaningful financial topics. This circularity is load-bearing because topic-precision is then used to weight NPMI coherence and intra/intertopic similarity in Tables 4-6. The unweighted NPMI coherence reported in parentheses in Table 4 actually favors AM (0.341 vs. 0.279 for sentences; 0.376 vs. 0.301 for refined sentences), so the standard metric contradicts the paper's topic-quality conclusion. The authors' dismissal of raw coherence in §4.2 is not justified by evidence, and it cannot compensate for the fact that the reported advantage vanishes when the evaluation is decoupled from the keyword list.
  2. [§3.3 and §3.4; Tables 2, 5, and 6] The comparison between FinTextSim and AM is confounded by the base model. FinTextSim is fine-tuned from ModernBERT, a recent, larger encoder model, while AM is all-MiniLM-L6-v2, an older and substantially smaller distilled model. The statement in §1 and §3.4 that 'keeping all other model parameters identical' isolates the effect of the sentence transformer is inaccurate, because the models differ in architecture, parameter count, training data, and release date. Without an untuned ModernBERT baseline or a fine-tuned AM model, the reported improvements in intratopic similarity, intertopic similarity, and outlier counts cannot be attributed to financial fine-tuning rather than to general superiority of the base architecture. This confound affects even the non-circular similarity metrics, so the central conclusion that 'domain-specific fine-tuning is crucial' is not established.
  3. [§3.4, c-tfidf seed words] The use of seed words with a multiplier of 50 in c-tfidf directly injects the evaluation keyword list into BERTopic's topic-word extraction. Since topic-precision is computed by matching topic words against that same list, this step further guarantees high topic-precision for FinTextSim, independently of the embedding quality. The paper does not report any sensitivity analysis for the seed multiplier or for the refined-sentence selection rule, both of which are free parameters that could drive the reported results. A comparison without seed words, or with an independently constructed evaluation vocabulary, is needed to support the claim that FinTextSim improves topic quality.
minor comments (4)
  1. [Abstract and §4.1, Table 2] The abstract claims that FinTextSim 'reduces intertopic similarity by 100%', but Table 2 reports 0.0002 vs. 0.4647, which is a 99.96% reduction. In addition, §4.1 states the FinTextSim intertopic similarity is '0.002', while Table 2 reports 0.0002; this inconsistency should be corrected.
  2. [Appendix A, Figure A.9 caption] Figure A.9 is captioned 'Wordcloud - FinTextSim - Refined Sentences', but given the sequence of Figures A.6-A.9, it appears the caption should read 'AM - Refined Sentences'.
  3. [§3.3] The fine-tuning procedure for FinTextSim does not report several hyperparameters needed for reproducibility, such as the number of training epochs, learning rate, and the exact ModernBERT variant used. The adaptive circle loss description also omits the pairing strategy for positive and negative sentence pairs.
  4. [§4.4] The sentence 'We find that BERTopic is highly on financial text when combined with FinTextSim' appears incomplete; it should probably read 'highly effective on financial text'.

Circularity Check

3 steps flagged · score 8.0 of 10

FinTextSim's 'perfect' topic-precision is a self-scoring loop: the same keyword list creates its training labels, seeds BERTopic's topic words, and defines the evaluation metric.

  1. self definitional [Section 3.4 'Model Creation' and Section 3.5.3 'Applicability for the Financial Domain' (Eq. 1, Table 3)]
    "To extract relevant economic topics, we use c-tfidf weighting, reduce common words, and guide the model by incorporating seed words based on the keyword list with a multiplier of 50. ... To evaluate the topic models specifically for the financial domain, we incorporate a topic-precision-based weighting into the metrics for topic quality and organizing power, using the previously described keyword list (see Section 3.2)."

    The same 14-topic keyword list is used to (i) create FinTextSim's training labels, (ii) retain refined sentences, (iii) seed BERTopic's c-tfidf with a multiplier of 50, and (iv) score the resulting topics via topic-precision (Eq. 1). Since topic-precision counts keywords from that list in BERTopic's topic representations, and those representations were boosted toward exactly those keywords, FinTextSim's perfect 1.0 in Table 3 measures reproduction of the model's own training/seed vocabulary rather than independently verified topic quality. The central detection result is forced by the experimental setup.

  2. fitted input called prediction [Section 3.3 'FinTextSim' and Section 3.5.2 'Organizing Power' (Table 2, Abstract)]
    "To tailor FinTextSim for financial text, we create a labeled dataset by adopting a dictionary-based approach, relying on the keyword list presented in Section 3.2. ... For the evaluation of FinTextSim, we use the topic labels from the test dataset."

    The headline improvements (intratopic similarity +81%, intertopic similarity -100%) are computed on the test split of a dataset labeled by the same keyword-dictionary rule used to train FinTextSim with adaptive circle loss. The test labels are generated by the same substring keyword matching procedure as the training labels, so the model is evaluated on reproducing the exact criterion it was optimized to satisfy. This is an in-distribution fit report for the dictionary labels, not evidence that FinTextSim discovers independently defined economic topics.

1 more flagged steps
  1. self definitional [Section 4.2 'Topic Quality' (Tables 3 and 4)]
    "Contrary to our expectations, BERTopic achieves higher coherence with AM than with FinTextSim when topic-precision-based weighting is not applied. ... Without domain-specific weighting, coherence fails to reflect practical applicability, making topic-precision essential for capturing meaningful financial insights."

    The paper explicitly concedes that unweighted NPMI coherence favors AM. The only mechanism that flips the topic-quality comparison in FinTextSim's favor is multiplication by topic-precision, which is defined from the same keyword list used to create FinTextSim's training labels and BERTopic's seed words. Weighted coherence therefore inherits the circularity: the reported 'financial topic quality' is a measure of keyword-list reproduction, not semantic coherence. The raw coherence result actually undercuts the claim that FinTextSim produces qualitatively superior topics.

full rationale

The circularity is concentrated in the evaluation design. FinTextSim is trained on sentences labeled by the authors' 14-topic keyword list (Section 3.3), BERTopic is guided with seed words from that same list at multiplier 50 (Section 3.4), and the headline metric topic-precision (Eq. 1) counts occurrences of keywords from that same list in the topic representations (Section 3.5.3). Tables 2 and 3 therefore report the model's fidelity to its own construction: a perfect 1.0 precision is the expected outcome, not an empirical discovery. The same circularity infects the 'weighted' coherence and similarity results in Tables 4-6, and the paper explicitly says that only this weighting converts FinTextSim's raw NPMI disadvantage into an advantage. There are some non-circular fragments — unweighted intra/intertopic similarity (parentheses in Tables 5-6) and the qualitative examples — but they are not the evidence for the central 'only FinTextSim yields clear clusters' claim. I do not count the ModernBERT-versus-MiniLM base-model confound as circularity, and there is no load-bearing self-citation chain. Because the primary quantitative claims reduce by construction to the keyword-list definition, the circularity score is 8 rather than 10: the paper does contain heldout test data and unweighted metrics, so the derivation is not fully equivalent to its input.

Assumptions & free parameters 9 free parameters · 5 assumptions · 1 invented entities

The central claim depends on a hand-built 14-topic keyword list used simultaneously for (1) labeling FinTextSim training data, (2) guiding BERTopic with seed words, and (3) computing the topic-precision metric that dominates the evaluation. This triple use makes the reported advantages largely self-fulfilling. No model or code is released, and the standard coherence metric points the other way.

free parameters (9)
  • Financial topic keyword list (14 topics) = 14 topics; excerpt in Appendix E
    Hand-constructed from Li (2010a) and Fengler & Phan (2023) with additions (Energy, ESG, Covid-19); used for training labels, seed words, and the topic-precision evaluation metric, so it directly determines the main result.
  • Sentence length filter for labeling = 5 to 50 words
    Thresholds chosen by hand in Section 3.3 to exclude noisy sentences; changes the composition of the FinTextSim training set.
  • Refined-sentence selection rule = keyword sentence plus pre/succeeding sentences
    Section 3.4: retains 678,218 of 687,959 sentences; defines the BERTopic input and therefore affects topic-precision and cluster metrics.
  • HDBSCAN min_cluster_size = 1,250
    Hand-set in Section 3.4 to prioritize global topics; strongly controls the number and sizes of clusters, hence all downstream topic metrics.
  • HDBSCAN min_samples = 10
    Section 3.4: controls outlier fraction; authors note higher values produce more noise. 184,470 outliers still remain with FinTextSim.
  • UMAP n_neighbors / min_dist / n_components = 100 / 0 / 10
    Section 3.3-3.4: hand-set to preserve global structure; affects cluster separability before HDBSCAN.
  • Seed word multiplier in c-tfidf = 50
    Section 3.4: biases BERTopic topic representations toward the keyword list, inflating topic-precision.
  • NPMI window size = 20
    Section 3.5.1: adjusted upward from the default 10 based on sentence length; affects raw coherence scores.
  • Document filtering thresholds = 250-word minimum, z-score > 2, low-cosine filter (threshold not given for main dataset)
    Section 3.1, Table 1: removes 3,161 of 4,600 filings (69%); the cosine threshold is not specified for the main dataset and may bias the corpus.
assumptions (5)
  • ad hoc to paper The 14-topic keyword list is a complete and accurate taxonomy of Item 7 and 7A topics, and a sentence containing two or more keywords from one domain is a valid single-topic training label.
    Invoked in Sections 3.2-3.3 to create FinTextSim labels and in Section 3.5.3 to score topic-precision; no human annotation or external validation is provided.
  • ad hoc to paper Topic-precision (Eq. 1) is a valid adjustment to NPMI coherence and intra/intertopic similarity for financial text.
    Section 3.5.3: the weighting reverses the raw NPMI ordering (Table 4) and favors the model trained on the same keyword list; its validity is asserted, not demonstrated.
  • domain assumption BERTopic's assumptions hold for sentence-level financial text: each sentence belongs to one topic, and UMAP/HDBSCAN clusters correspond to semantic topics.
    Acknowledged in Sections 2.2 and 3.4; the paper uses sentence-level input to satisfy the single-topic assumption but does not test whether the clusters are stable or meaningful.
  • domain assumption ModernBERT with mean pooling and adaptive circle loss is an appropriate base configuration for financial sentence embeddings.
    Section 3.3: this is a modeling choice from prior literature (Warner et al. 2024; Sun et al. 2020); the paper provides no ablation or comparison with other base models.
  • domain assumption The external S&P 500 annual and sustainability report dataset is distributionally compatible with 10-K Item 7/7A text for keyword-based labeling.
    Section 3.3: used to expand training data; the distribution shift from sustainability reports to MD&A text is not analyzed.
invented entities (1)
  • FinTextSim
    purpose: Domain-specific sentence transformer for clustering and semantic search in financial text.
    No model weights, training code, or full training data are released; performance evidence comes only from the paper's circular evaluation, so there is no external falsifiable handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FinTextSim: Enhancing Financial Text Analysis with BERTopic." pith.science (2026). https://pith.science/paper/BHIXCSVP

@misc{pith2026250415683,
  author       = {Pith},
  title        = {Pith review of: FinTextSim: Enhancing Financial Text Analysis with BERTopic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BHIXCSVP}},
  note         = {Machine review of arXiv:2504.15683}
}
read the original abstract

Recent advancements in information availability and computational capabilities have transformed the analysis of annual reports, integrating traditional financial metrics with insights from textual data. To extract valuable insights from this wealth of textual data, automated review processes, such as topic modeling, are crucial. This study examines the effectiveness of BERTopic, a state-of-the-art topic model relying on contextual embeddings, for analyzing Item 7 and Item 7A of 10-K filings from S&P 500 companies (2016-2022). Moreover, we introduce FinTextSim, a finetuned sentence-transformer model optimized for clustering and semantic search in financial contexts. Compared to all-MiniLM-L6-v2, the most widely used sentence-transformer, FinTextSim increases intratopic similarity by 81% and reduces intertopic similarity by 100%, significantly enhancing organizational clarity. We assess BERTopic's performance using embeddings from both FinTextSim and all-MiniLM-L6-v2. Our findings reveal that BERTopic only forms clear and distinct economic topic clusters when paired with FinTextSim's embeddings. Without FinTextSim, BERTopic struggles with misclassification and overlapping topics. Thus, FinTextSim is pivotal for advancing financial text analysis. FinTextSim's enhanced contextual embeddings, tailored for the financial domain, elevate the quality of future research and financial information. This improved quality of financial information will enable stakeholders to gain a competitive advantage, streamlining resource allocation and decision-making processes. Moreover, the improved insights have the potential to leverage business valuation and stock price prediction models.

Figures

Figures reproduced from arXiv: 2504.15683 by the authors.

Figure 1
Figure 1. FinTextSim vs. AM on the test dataset. The colors of the datapoints represent a topic from the keyword list. contrast, AM fails to accurately detect topic specifics, resulting in an indis￾tinguishable mash of data points (see [PITH_FULL_IMAGE:figures/full_fig_p018_1.png] view at source ↗
Figure 2
Figure 2. Topic representations - FinTextSim vs. AM - HR. Original cleaned sentence: ’a majority of employees belong to labor unions’. Turning to a practical example, BERTopic with FinTextSim is able to accurately capture the underlying topic of a sentence, while the results with AM are misleading [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Topic representations - FinTextSim vs. AM - Cost. Original cleaned sentence: ’business is vulnerable to fluctuations in fuel costs and disruptions in fuel supplies’. To illustrate this in a practical example, we refer to [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Topic representations - FinTextSim vs. AM - Operations. Original cleaned sentence: ’the company manufactures markets and distributes spices seasoning mixes condiments and other flavorful products to the entire food industry retailers food manufacturers and foodservice …
Figure 5
Figure 5. Figure 5: Topic representations - FinTextSim vs. AM - Accounting. Original cleaned sentence: ’critical accounting policies estimates and judgments our consolidated financial statements are based on gaap which requires us to make estimates and assumptions about future events that…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

91 extracted references · 61 canonical work pages

  1. [1]

    , author Eid, Y

    author Abdelrazek, A. , author Eid, Y. , author Gawish, E. , author Medhat, W. , author Hassan, A. , year 2023 . title Topic modeling algorithms and applications: A survey . journal Information Systems volume 112 , pages 102131

    2023

  2. [2]

    , author Dutta, S

    author Abukari, K. , author Dutta, S. , author Li, C. , author Tang, S. , author Zhu, P. , year 2024 . title Corporate communication and likelihood of data breaches . journal International Review of Economics & Finance volume 94 , pages 103433 . https://www.sciencedirect.com/science/article/pii/S1059056024004258, :https://doi.org/10.1016/j.iref.2024.103433

    arXiv 2024

  3. [3]

    , author Al-Khalifa, H

    author Abuzayed, A. , author Al-Khalifa, H. , year 2021 . title Bert for arabic topic modeling: An experimental study on bertopic technique . journal Procedia computer science volume 189 , pages 191--194

    2021

  4. [4]

    , author Fu, W

    author Agrawal, A. , author Fu, W. , author Menzies, T. , year 2018 . title What is wrong with topic modeling? and how to fix it using search-based software engineering . journal Information and Software Technology volume 98 , pages 74--88

    2018

  5. [5]

    , author Yeap, T.H

    author Albalawi, R. , author Yeap, T.H. , author Benyoucef, M. , year 2020 . title Using topic modeling methods for short-text data: A comparative analysis . journal Frontiers in artificial intelligence volume 3 , pages 42

    2020

  6. [6]

    , author Kherfi, M.L

    author Allaoui, M. , author Kherfi, M.L. , author Cheriet, A. , year 2020 . title Considerably improving clustering algorithms using umap dimensionality reduction technique: a comparative study , in: booktitle International conference on image and signal processing , organization Springer . pp. pages 317--325

    2020

  7. [7]

    , author Bone, C

    author Altaweel, M. , author Bone, C. , author Abrams, J. , year 2019 . title Documents as data: a content analysis and topic modeling approach for analyzing responses to ecological disturbances . journal Ecological Informatics volume 51 , pages 82--95

    2019

  8. [8]

    , year 2020

    author Angelov, D. , year 2020 . title Top2vec: Distributed representations of topics . journal arXiv preprint arXiv:2008.09470

    arXiv 2020

Show all 91 references
  1. [9]

    , year 2019

    author Araci, D. , year 2019 . title Finbert: Financial sentiment analysis with pre-trained language models . journal arXiv preprint arXiv:1908.10063

    2019 arXiv

  2. [10]

    , author Pipal, C

    author Baden, C. , author Pipal, C. , author Schoonvelde, M. , author van der Velden, M.A.G. , year 2022 . title Three gaps in computational text analysis methods for social sciences: A research agenda . journal Communication Methods and Measures volume 16 , pages 1--18

    2022

  3. [11]

    , author Datta, A

    author Bao, Y. , author Datta, A. , year 2014 . title Simultaneously discovering and quantifying risk types from textual risk disclosures . journal Management Science volume 60 , pages 1371--1391

    2014

  4. [12]

    , author Bhagat, S

    author Bellstam, G. , author Bhagat, S. , author Cookson, J.A. , year 2021 . title A text-based analysis of corporate innovation . journal Management Science volume 67 , pages 4004--4031

    2021

  5. [13]

    , author Mickovic, A

    author Bhattacharya, I. , author Mickovic, A. , year 2024 . title Accounting fraud detection using contextual language learning . journal International Journal of Accounting Information Systems volume 53 , pages 100682

    2024

  6. [14]

    , author Bi, Y

    author Blair, S.J. , author Bi, Y. , author Mulvenna, M.D. , year 2020 . title Aggregated topic models for increasing social media topic coherence . journal Applied Intelligence volume 50 , pages 138--156

    2020

  7. [15]

    , author Ng, A.Y

    author Blei, D.M. , author Ng, A.Y. , author Jordan, M.I. , year 2003 . title Latent dirichlet allocation . journal Journal of machine Learning research volume 3 , pages 993--1022

    2003

  8. [16]

    , author Chiu, V

    author Booker, A. , author Chiu, V. , author Groff, N. , author Richardson, V.J. , year 2024 . title Ais research opportunities utilizing machine learning: From a meta-theory of accounting literature . journal International Journal of Accounting Information Systems volume 52 ,...

    2024

  9. [17]

    , author Cortijo-Gallego, V

    author Borrero-Dom \' nguez, C. , author Cortijo-Gallego, V. , author Escobar-Rodr \' guez, T. , year 2024 . title Digital transformation voluntary disclosure: Insights from leading european companies . journal International Journal of Accounting Information Systems volume 55 ...

    2024

  10. [18]

    , author Crowley, R.M

    author Brown, N.C. , author Crowley, R.M. , author Elliott, W.B. , year 2020 . title What are you saying? using topic to detect financial misreporting . journal Journal of Accounting Research volume 58 , pages 237--291

    2020

  11. [19]

    , author Hanley, K.W

    author Cai, K.N. , author Hanley, K.W. , author Huang, A.G. , author Zhao, X. , year 2022 . title Risk disclosure and the pricing of corporate debt issues in private and public markets . journal Georgetown McDonough School of Business Research Paper

    2022

  12. [20]

    , author Hindle, A

    author Campbell, J.C. , author Hindle, A. , author Stroulia, E. , year 2015 . title Latent dirichlet allocation: extracting topics from software engineering data , in: booktitle The art and science of analyzing software data . publisher Elsevier , pp. pages 139--159

    2015

  13. [21]

    , author Fontana, G

    author Canelli, R. , author Fontana, G. , author Realfonzo, R. , author Passarella, M.V. , year 2024 . title Energy crisis, economic growth and public finance in italy . journal Energy Economics volume 132 , pages 107430

    2024

  14. [22]

    , author M \"u ller, O

    author Caron, M. , author M \"u ller, O. , year 2020 . title Hardening soft information: A transformer-based approach to forecasting stock return volatility , in: booktitle 2020 IEEE International Conference on Big Data (Big Data) , organization IEEE . pp. pages 4383--4391

    2020

  15. [23]

    , author Rabbani, R.M

    author Chen, Y. , author Rabbani, R.M. , author Gupta, A. , author Zaki, M.J. , year 2017 . title Comparative text analytics via topic modeling in banking , in: booktitle 2017 IEEE symposium series on computational intelligence (SSCI) , organization IEEE . pp. pages 1--8

    2017

  16. [24]

    , author Zhang, H

    author Chen, Y. , author Zhang, H. , author Liu, R. , author Ye, Z. , author Lin, J. , year 2019 . title Experimental explorations on short text topic mining between lda and nmf based schemes . journal Knowledge-Based Systems volume 163 , pages 1--13

    2019

  17. [25]

    , author Malloy, C

    author Cohen, L. , author Malloy, C. , author Nguyen, Q. , year 2020 . title Lazy prices . journal The Journal of Finance volume 75 , pages 1371--1415

    2020

  18. [26]

    , author Kadach, I

    author Cohen, S. , author Kadach, I. , author Ormazabal, G. , author Reichelstein, S. , year 2023 . title Executive compensation tied to esg performance: International evidence . journal Journal of Accounting Research volume 61 , pages 805--853

    2023

  19. [27]

    , author Erhard, R.D

    author Dechow, P.M. , author Erhard, R.D. , author Sloan, R.G. , author Soliman, M.T. , year 2021 . title Implied equity duration: A measure of pandemic shutdown risk . journal Journal of Accounting Research volume 59 , pages 243--281

    2021

  20. [28]

    , author Dumais, S.T

    author Deerwester, S. , author Dumais, S.T. , author Furnas, G.W. , author Landauer, T.K. , author Harshman, R. , year 1990 . title Indexing by latent semantic analysis . journal Journal of the American society for information science volume 41 , pages 391--407

    1990

  21. [29]

    , author Chang, M.W

    author Devlin, J. , author Chang, M.W. , author Lee, K. , author Toutanova, K. , year 2019 . title Bert: Pre-training of deep bidirectional transformers for language understanding. arxiv . journal arXiv preprint arXiv:1810.04805

    2019 arXiv

  22. [30]

    , author Stratopoulos, T.C

    author Dong, M.M. , author Stratopoulos, T.C. , author Wang, V.X. , year 2024 . title A scoping review of chatgpt research in accounting and finance . journal International Journal of Accounting Information Systems volume 55 , pages 100715

    2024

  23. [31]

    , author Stodden, V

    author Donoho, D. , author Stodden, V. , year 2003 . title When does non-negative matrix factorization give a correct decomposition into parts? journal Advances in neural information processing systems volume 16

    2003

  24. [32]

    , author Fuksa, M

    author Dutta, S. , author Fuksa, M. , author Macaulay, K. , year 2019 . title Determinants of md&a sentiment in canada . journal International Review of Economics & Finance volume 60 , pages 130--148 . https://www.sciencedirect.com/science/article/pii/S105905601630332X, :https...

    2019 doi

  25. [33]

    , author Lang, M

    author Dyer, T. , author Lang, M. , author Stice-Lawrence, L. , year 2017 . title The evolution of 10-k textual disclosure: Evidence from latent dirichlet allocation . journal Journal of Accounting and Economics volume 64 , pages 221--245

    2017

  26. [34]

    , author Yu, J

    author Egger, R. , author Yu, J. , year 2021 . title Identifying hidden semantic structures in instagram data: a topic modelling comparison . journal Tourism Review volume 77 , pages 1234--1246

    2021

  27. [35]

    , author Yu, J

    author Egger, R. , author Yu, J. , year 2022 . title A topic modeling comparison between lda, nmf, top2vec, and bertopic to demystify twitter posts . journal Frontiers in sociology volume 7 , pages 886498

    2022

  28. [36]

    , author Vanani, I.R

    author Farzadnia, S. , author Vanani, I.R. , author Hanafizadeh, P. , year 2024 . title An experimental study for identifying customer prominent viewpoints on different flight classes by topic modeling methods . journal International Journal of Information Management Data Insi...

    2024

  29. [37]

    , author Phan, M.T

    author Fengler, M. , author Phan, M.T. , year 2023 . title A Topic Model for 10-K Management Disclosures . type Economics Working Paper Series number 2307 . University of St. Gallen, School of Economics and Political Science. https://EconPapers.repec.org/RePEc:usg:econwp:2023:07

    2023

  30. [38]

    , author Gkolia, A

    author Fernandes, N. , author Gkolia, A. , author Pizzo, N. , author Davenport, J. , author Nair, A. , year 2020 . title Unification of hdp and lda models for optimal topic clustering of subject specific question banks . journal arXiv preprint arXiv:2011.01035

    2020 arXiv

  31. [39]

    , author Gazzani, A

    author Ferriani, F. , author Gazzani, A. , year 2023 . title The invasion of ukraine and the energy crisis: Comparative advantages in equity valuations . journal Finance Research Letters volume 58 , pages 104604

    2023

  32. [40]

    , author Zhuang, Y

    author Fu, Q. , author Zhuang, Y. , author Gu, J. , author Zhu, Y. , author Guo, X. , year 2021 . title Agreeing to disagree: Choosing among eight topic-modeling methods . journal Big Data Research volume 23 , pages 100173

    2021

  33. [41]

    , author Nair, V

    author Fu, W. , author Nair, V. , author Menzies, T. , year 2016 . title Why is differential evolution better than grid search for tuning defect predictors? journal arXiv preprint arXiv:1609.02613

    2016 arXiv

  34. [42]

    , author Liu, S

    author Gao, Y. , author Liu, S. , author Yang, L. , year 2025 . title Artificial intelligence and innovation capability: A dynamic capabilities perspective . journal International Review of Economics & Finance volume 98 , pages 103923 . https://www.sciencedirect.com/science/ar...

    2025

  35. [43]

    , author de Arriba-P\'erez, F

    author Garc\'ia-M\'endez, S. , author de Arriba-P\'erez, F. , author Barros-Vila, A. , author Gonz\'alez-Casta\ no, F.J. , author Costa-Montenegro, E. , year 2023 . title Automatic detection of relevant information, predictions and forecasts in financial news through topic mod...

    2023

  36. [44]

    , author Vavasis, S.A

    author Gillis, N. , author Vavasis, S.A. , year 2014 . title Fast and Robust Recursive Algorithms for Separable Nonnegative Matrix Factorization . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 36 , pages 698--714 . :10.1109/TPAMI.2013.226, http:...

    2014 arXiv

  37. [45]

    , year 2003

    author Griffin, P.A. , year 2003 . title Got information? investor response to form 10-k and form 10-q edgar filings . journal Review of Accounting Studies volume 8 , pages 433--460

    2003

  38. [46]

    , author Pintilie, I

    author Grigore, D.N. , author Pintilie, I. , year 2023 . title Transformer-based topic modeling to measure the severity of eating disorder symptoms. , in: booktitle CLEF (Working Notes) , pp. pages 684--692

    2023

  39. [47]

    , year 2022

    author Grootendorst, M. , year 2022 . title Bertopic: Neural topic modeling with a class-based tf-idf procedure . journal arXiv preprint arXiv:2203.05794

    2022 arXiv

  40. [48]

    , author Schreyer, M

    author Gu, H. , author Schreyer, M. , author Moffitt, K. , author Vasarhelyi, M. , year 2024 a. title Artificial intelligence co-piloted auditing . journal International Journal of Accounting Information Systems volume 54 , pages 100698

    2024

  41. [49]

    , author Dai, J

    author Gu, Y. , author Dai, J. , author Vasarhelyi, M.A. , year 2023 . title Audit 4.0-based esg assurance: An example of using satellite images on ghg emissions . journal International Journal of Accounting Information Systems volume 50 , pages 100625

    2023

  42. [50]

    , author Katz, S

    author Gu, Y. , author Katz, S. , author Wang, X. , author Vasarhelyi, M. , author Dai, J. , year 2024 b. title Government esg reporting in smart cities . journal International Journal of Accounting Information Systems volume 54 , pages 100701

    2024

  43. [51]

    , author Zong, X

    author Guo, M. , author Zong, X. , author Guo, L. , author Lei, Y. , year 2024 . title Does haze-related sentiment affect income inequality in china? journal International Review of Economics & Finance volume 94 , pages 103371 . https://www.sciencedirect.com/science/article/pi...

    2024 doi

  44. [52]

    , author Dengre, V

    author Gupta, A. , author Dengre, V. , author Kheruwala, H.A. , author Shah, M. , year 2020 . title Comprehensive review of text-mining applications in finance . journal Financial Innovation volume 6 , pages 1--25

    2020

  45. [53]

    , author Guo, W

    author Han, D. , author Guo, W. , author Chen, H. , author Wang, B. , author Guo, Z. , year 2024 . title Lest: Large language models and spatio-temporal data analysis for enhanced sino-us exchange rate forecasting . journal International Review of Economics & Finance volume 96...

    2024

  46. [54]

    , author Zheng, X

    author Hong, W. , author Zheng, X. , author Qi, J. , author Wang, W. , author Zheng, N. , author Weng, Y. , year 2018 . title Financialflow: Visual analytics of financial news based on hierarchical dirichlet process , in: booktitle 2018 33rd Youth Academic Annual Conference of...

    2018

  47. [55]

    , author Hristova, D

    author Hsieh, H.T. , author Hristova, D. , year 2022 . title Transformer-based summarization and sentiment analysis of sec 10-k annual reports for company performance prediction , in: booktitle Proceedings of the 55th Hawaii International Conference on System Sciences , publis...

    2022 doi

  48. [56]

    , author Wang, H

    author Huang, A.H. , author Wang, H. , author Yang, Y. , year 2023 . title Finbert: A large language model for extracting information from financial text . journal Contemporary Accounting Research volume 40 , pages 806--841

    2023

  49. [57]

    , author Wu, D.A

    author Jegadeesh, N. , author Wu, D.A. , year 2017 . title Deciphering Fedspeak : The Information Content of FOMC Meetings . journal SSRN Electronic Journal :10.2139/ssrn.2939937

    2017 doi

  50. [58]

    , author Ghodsi, A

    author Karami, M. , author Ghodsi, A. , year 2024 . title Orchid: Flexible and data-dependent convolution for sequence modeling . journal arXiv preprint arXiv:2402.18508

    2024

  51. [59]

    , author Kim, K.S

    author Kim, M.G. , author Kim, K.S. , author Lee, K.C. , year 2022 . title Analyzing the effects of topics underlying companies' financial disclosures about risk factors on prediction of esg risk ratings: Emphasis on bertopic , in: booktitle 2022 IEEE International Conference ...

    2022

  52. [60]

    , author Seung, H.S

    author Lee, D.D. , author Seung, H.S. , year 1999 . title Learning the parts of objects by non-negative matrix factorization . journal Nature volume 401 , pages 788--791

    1999

  53. [61]

    , year 2010 a

    author Li, F. , year 2010 a. title The Information Content of Forward - Looking Statements in Corporate Filings --- A Na \"i ve Bayesian Machine Learning Approach . journal Journal of Accounting Research volume 48 , pages 1049--1102 . :10.1111/j.1475-679X.2010.00382.x

    2010 arXiv

  54. [62]

    , year 2010 b

    author Li, F. , year 2010 b. title Textual analysis of corporate disclosures: A survey of the literature . journal Journal of Accounting Literature volume 29 , pages 143--165

    2010

  55. [63]

    , author Chen, H

    author Li, T. , author Chen, H. , author Liu, W. , author Yu, G. , author Yu, Y. , year 2023 . title Understanding the role of social media sentiment in identifying irrational herding behavior in the stock market . journal International Review of Economics & Finance volume 87 ...

    2023 doi

  56. [64]

    , author Hwang, Y

    author Lin, H. , author Hwang, Y. , year 2021 . title The effects of personal information management capabilities and social-psychological factors on accounting professionals’ knowledge-sharing intentions: Pre and post covid-19 . journal International Journal of Accounting Inf...

    2021

  57. [65]

    , year 2022

    author Liu, M. , year 2022 . title Assessing human information processing in lending decisions: A machine learning approach . journal Journal of Accounting Research volume 60 , pages 607--651

    2022

  58. [66]

    , author McDonald, B

    author Loughran, T. , author McDonald, B. , year 2011 . title When is a liability not a liability? textual analysis, dictionaries, and 10-ks . journal The Journal of finance volume 66 , pages 35--65

    2011

  59. [67]

    , author Michaely, R

    author Lowry, M. , author Michaely, R. , author Volkova, E. , year 2020 . title Information revealed through the regulatory process: Interactions between the sec and companies ahead of their ipo . journal The Review of Financial Studies volume 33 , pages 5510--5554

    2020

  60. [68]

    , year 2022

    author Lu, J. , year 2022 . title Limited attention: Implications for financial reporting . journal Journal of Accounting Research volume 60 , pages 1991--2027

    2022

  61. [69]

    , author Waldherr, A

    author Maier, D. , author Waldherr, A. , author Miltner, P. , author Wiedemann, G. , author Niekler, A. , author Keinert, A. , author Pfetsch, B. , author Heyer, G. , author Reber, U. , author H \"a ussler, T. , author Schmid-Petri , H. , author Adam, S. , year 2018 . title Ap...

    2018

  62. [70]

    , author Paroubek, P

    author Masson, C. , author Paroubek, P. , year 2020 . title Nlp analytics in finance with dore: a french 257m tokens corpus of corporate annual reports , in: booktitle Language Resources and Evaluation Conference (LREC 2020) , organization ELRA . pp. pages 2261--2267

    2020

  63. [71]

    , author Fithriasari, K

    author Mayasari, R.W. , author Fithriasari, K. , author Prastyo, D.D. , year 2021 . title Text mining to analyse publication topics of covid-19 using hdp and lda methods , in: booktitle AECon 2020: Proceedings of The 6th Asia-Pacific Education And Science Conference, AECon 202...

    2021

  64. [72]

    , author Healy, J

    author McInnes, L. , author Healy, J. , year 2017 . title Accelerated Hierarchical Density Clustering , in: booktitle 2017 IEEE International Conference on Data Mining Workshops ( ICDMW ) , pp. pages 33--42 . :10.1109/ICDMW.2017.12, http://arxiv.org/abs/1705.07321 arXiv:1705.07321

    2017 arXiv

  65. [73]

    , author Feeney, O

    author Murphy, B. , author Feeney, O. , author Rosati, P. , author Lynn, T. , year 2024 . title Exploring accounting and ai using topic modelling . journal International Journal of Accounting Information Systems volume 55 , pages 100709

    2024

  66. [74]

    , author Greene, D

    author O'Callaghan, D. , author Greene, D. , author Carthy, J. , author Cunningham, P. , year 2015 . title An analysis of the coherence of descriptors in topic modeling . journal Expert Systems with Applications volume 42 , pages 5645--5657 . :10.1016/j.eswa.2015.02.055

    2015 doi

  67. [75]

    a rvenp \

    author Ranta, M. , author Ylinen, M. , author J \"a rvenp \"a \"a , M. , year 2022 . title Machine Learning in Management Accounting Research : Literature Review and Pathways for the Future . journal European Accounting Review , pages 1--30 :10.1080/09638180.2022.2137221

    2022

  68. [76]

    , author Gurevych, I

    author Reimers, N. , author Gurevych, I. , year 2019 . title Sentence-bert: Sentence embeddings using siamese bert-networks . journal arXiv preprint arXiv:1908.10084

    2019 arXiv

  69. [77]

    , author Both, A

    author R \"o der, M. , author Both, A. , author Hinneburg, A. , year 2015 . title Exploring the space of topic coherence measures , in: booktitle Proceedings of the eighth ACM international conference on Web search and data mining , pp. pages 399--408

    2015

  70. [78]

    , author Rey-Moreno, M

    author S \'a nchez-Franco, M.J. , author Rey-Moreno, M. , year 2022 . title Do travelers' reviews depend on the destination? an analysis in coastal and urban peer-to-peer lodgings . journal Psychology & marketing volume 39 , pages 441--459

    2022

  71. [79]

    , author Dalmia, A

    author Sia, S. , author Dalmia, A. , author Mielke, S.J. , year 2020 . title Tired of Topic Models ? Clusters of Pretrained Word Embeddings Make for Fast and Good Topics too! , in: booktitle Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing...

    2020 doi

  72. [80]

    , author Cheng, C

    author Sun, Y. , author Cheng, C. , author Zhang, Y. , author Zhang, C. , author Zheng, L. , author Wang, Z. , author Wei, Y. , year 2020 . title Circle loss: A unified perspective of pair similarity optimization , in: booktitle Proceedings of the IEEE/CVF conference on comput...

    2020

  73. [81]

    , author Jordan, M.I

    author Teh, Y.W. , author Jordan, M.I. , author Beal, M.J. , author work(s): , D.M.B.R. , year 2006 . title Hierarchical Dirichlet Processes . journal Journal of the American Statistical Association volume 101 , pages 1566--1581 . http://arxiv.org/abs/27639773 arXiv:27639773

    2006

  74. [82]

    , author Shazeer, N

    author Vaswani, A. , author Shazeer, N. , author Parmar, N. , author Uszkoreit, J. , author Jones, L. , author Gomez, A.N. , author Kaiser, L. , author Polosukhin, I. , year 2017 . title Attention Is All You Need , in: booktitle 31st Conference on Neural Information Processing...

    2017 arXiv

  75. [83]

    , author Zhang, X.L

    author Wang, J. , author Zhang, X.L. , year 2023 . title Deep nmf topic modeling . journal Neurocomputing volume 515 , pages 157--173

    2023

  76. [84]

    , author Chaffin, A

    author Warner, B. , author Chaffin, A. , author Clavi \'e , B. , author Weller, O. , author Hallstr \"o m, O. , author Taghadouini, S. , author Gallagher, A. , author Biswas, R. , author Ladhak, F. , author Aarsen, T. , et al., year 2024 . title Smarter, better, faster, longer...

    2024 arXiv

  77. [85]

    , author Nguyen, T

    author Wu, X. , author Nguyen, T. , author Luu, A.T. , year 2024 . title A survey on neural topic models: methods, applications, and challenges . journal Artificial Intelligence Review volume 57 , pages 18

    2024

  78. [86]

    , author Porter, A

    author Yau, C.K. , author Porter, A. , author Newman, N. , author Suominen, A. , year 2014 . title Clustering scientific documents with topic modeling . journal Scientometrics volume 100 , pages 767--786

    2014

  79. [87]

    , author Zhang, X.j

    author You, H. , author Zhang, X.j. , year 2009 . title Financial reporting complexity and investor underreaction to 10-k information . journal Review of Accounting studies volume 14 , pages 559--586

    2009

  80. [88]

    , author Phung, D

    author Zhao, H. , author Phung, D. , author Huynh, V. , author Jin, Y. , author Du, L. , author Buntine, W. , year 2021 . title Topic modelling meets deep neural networks: A survey . journal arXiv preprint arXiv:2103.00498

    2021 arXiv

  81. [89]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  82. [90]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  83. [91]

    @ A A A 1@ 0 ] A A A tA A A A D. c A A A a

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.