Triplet TF-IDF-like Measure in Text Analysis

• The triplet TF-IDF-like measure extends classical TF-IDF by incorporating three factors: local frequency, global rarity, and an additional supervised or distributional component.
• It is interpreted in two ways: as a three-factor decomposition (e.g., Term_Class Relevancy, TF-PI) and as an extension of the TF-IDF logic to triplet features.
• Empirical findings indicate improved classification accuracy in supervised settings, though challenges such as sparsity and model tuning remain.

A triplet TF-IDF-like measure is a weighting construction that preserves the basic TF-IDF logic of combining local prominence with global rarity, but introduces a third factor or extends the same logic from single terms to triplet features. The literature supports both readings. In supervised text categorization, the relevant pattern is a three-factor score such as Term_Class Relevancy or TF-PI, where the additional factor encodes class prior, class-certainty gain, or another label-aware quantity (Guru et al., 2016, Zhang, 2023). In significance-testing work, the same idea suggests replacing unigram counts by counts of a triplet feature and defining TF-IDF-, TF-ICF-, or Fisher-style enrichment scores over triplet opportunities rather than token positions (Sheridan et al., 21 Jul 2025).

1. Conceptual scope

Classical TF-IDF is a two-factor weighting scheme. In one standard formulation,

$\mathrm{TF\mbox{-}IDF}(t_i,d_j)=k_{ij}\times \log\!\left(\frac{N}{K_i}\right),$

where $k_{ij}$ is the number of times term $t_i$ occurs in document $d_j$ and $K_i$ is the number of documents containing at least one occurrence of $t_i$ (Sheridan et al., 2020). A related formulation used in statistical analyses is

$\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$

with $n_{ij}$ the term count in document $d_j$ and $b_i$ the document frequency (Sheridan et al., 21 Jul 2025).

The expression “triplet TF-IDF-like measure” therefore has two technically distinct meanings. In several supervised weighting schemes, “triplet” denotes a three-factor decomposition: a local term-frequency-like factor, a global rarity or specificity factor, and a third factor such as class prior, class-certainty gain, entropy-derived concentration, or burstiness correction (Guru et al., 2016, Zhang, 2023, Ahmed et al., 1 Apr 2026). In a different line of work, “triplet” refers to the feature itself, as in a triplet $k_{ij}$0, with the TF-IDF or Fisher-style logic transferred from unigrams to higher-order counted units (Sheridan et al., 21 Jul 2025).

A common misconception is to conflate these two meanings. The sources do not support that equivalence. A tri-factor weighting over unigrams is not the same object as a weighting defined on trigrams or other triplet features, even though both are reasonably described as triplet TF-IDF-like measures.

2. Three-factor supervised constructions

The clearest explicitly tri-factor construction is the Term_Class Relevancy measure introduced for text categorization. Its final score is

$k_{ij}$1

where $k_{ij}$2 is the proportion of documents containing $k_{ij}$3 that belong to class $k_{ij}$4, $k_{ij}$5 is the proportion of all occurrences of $k_{ij}$6 that lie inside class $k_{ij}$7, and $k_{ij}$8 is the proportion of documents in the corpus that belong to class $k_{ij}$9 (Guru et al., 2016). Although the abstract emphasizes only two factors, the full scoring function explicitly multiplies three components. Structurally, it is best interpreted as a supervised class-conditional composite term relevance measure rather than a plain TF-IDF replacement.

A second explicit three-factor construction is TF-PI, derived from the troenpy-based Positive Class Frequency weighting. The paper defines

$t_i$0

and then

$t_i$1

Here $t_i$2 is local term frequency, $t_i$3 is global inverse document frequency, and $t_i$4 is a supervised class-certainty gain computed from label distributions (Zhang, 2023). In that formulation, the third factor is not a prior but a label-aware commonness or certainty term derived from troenpy.

These two schemes illustrate the main tri-factor pattern already present in the literature. Both retain a multiplicative architecture analogous in spirit to TF-IDF, but both replace TF-IDF’s purely unsupervised document-centric logic with class-conditional information.

3. Factor taxonomy

Across the cited work, the three parts of a triplet TF-IDF-like measure fall into recurring categories.

The first category is the local factor. In classical text weighting this is raw term frequency, such as $t_i$5 or $t_i$6 (Sheridan et al., 2020, Sheridan et al., 21 Jul 2025). In scATAC-seq, the corresponding local quantity is

$t_i$7

where $t_i$8 after binarization, so the local component is a length-normalized binary incidence rather than a count intensity (Zandigohar et al., 2022).

The second category is the global rarity or specificity factor. Classical IDF takes the form $t_i$9 or $d_j$0 (Sheridan et al., 2020, Sheridan et al., 21 Jul 2025). In the Term_Class Relevancy framework, the analogous component is

$d_j$1

which uses document-frequency information but acts as a class-purity ratio rather than a rarity penalty (Guru et al., 2016). In scATAC-seq, the analogous global term is

$d_j$2

which downweights peaks accessible in many cells and upweights peaks that are rare across cells (Zandigohar et al., 2022).

The third category is an additional supervised or distributional factor. In Term_Class Relevancy this is the class prior

$d_j$3

In TF-PI it is the supervised certainty gain $d_j$4, defined through troenpy of class-label distributions (Zhang, 2023). In the scATAC-seq study, the entropy-based variant introduces a third global reweighting ingredient with

$d_j$5

and an entropy quantity involving

$d_j$6

combined with a log-scaled local factor $d_j$7 (Zandigohar et al., 2022). In the penalized likelihood-ratio framework for burstiness, the third component is a TF-like repetition term,

$d_j$8

appearing alongside TF-ICF and BTF-IDF inside the derived score (Ahmed et al., 1 Apr 2026).

This suggests a general three-part template: local evidence, global background rarity, and an additional factor that captures class bias, concentration, commonness, certainty, entropy, or burstiness. The sources do not present a single universal formalism, but they repeatedly instantiate that architecture.

4. Statistical interpretations and triplet-feature extensions

A separate line of work interprets TF-IDF-like weights as approximations to statistical significance scores. The hypergeometric-test study defines

$d_j$9

and reports that this negative log hypergeometric tail probability behaves very similarly in practice to TF-IDF on retrieval, summarization, and classification tasks (Sheridan et al., 2020). The statistical interpretation is over-representation: a term is important when its within-document concentration is unlikely under the corpus background.

The Fisher exact test justification sharpens that connection. For a unigram $K_i$0 in document $K_i$1, the one-tailed Fisher p-value $K_i$2 yields the significance weight $K_i$3, and the paper shows that under mild regularity conditions this is approximately TF-ICF plus correction terms; under stronger idealized assumptions, the same logic collapses to TF-IDF (Sheridan et al., 21 Jul 2025). The key closed forms are

$K_i$4

and

$K_i$5

That paper does not prove a theorem for triplets, but it explicitly suggests replacing the unigram $K_i$6 by a triplet feature $K_i$7. With triplet counts $K_i$8, total triplet count $K_i$9, document frequency $t_i$0, triplet opportunities $t_i$1, and total triplet opportunities $t_i$2, the direct analogues are

$t_i$3

and

$t_i$4

together with the Fisher-style significance score

$t_i$5

The source presents this as the most faithful extension of its derivation rather than as a proved result for triplets (Sheridan et al., 21 Jul 2025).

The penalized likelihood-ratio analysis of burstiness provides a different statistical route to a tri-factor interpretation. Its derived score contains

$t_i$6

$t_i$7

and

$t_i$8

plus document-length and collection-level burstiness corrections (Ahmed et al., 1 Apr 2026). In that framework, a triplet TF-IDF-like measure is not merely heuristic multiplication, but a multi-factor score arising from a formal test of non-burstiness versus burstiness.

5. Representation schemes and application domains

In supervised text categorization, the tri-factor weight is often not the final representation but an intermediate object. After computing Term_Class Relevancy for all terms, a document with $t_i$9 terms and $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$0 classes is mapped to a matrix $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$1 of size $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$2,

$\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$3

and then reduced to a $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$4-dimensional class-space vector

$\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$5

Thus each document is represented over classes rather than over the original vocabulary (Guru et al., 2016). The reported classifiers are SVM with Linear, RBF, and Polynomial kernels, and k-NN with Euclidean distance.

The scATAC-seq work shows that TF-IDF-like factorization is transferable beyond text. There the sparse binary matrix has rows as genomic regions or peaks and columns as cells, with the analogy

$\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$6

The main workflow is binary peak-by-cell matrix $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$7 TF-IDF transformation $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$8 truncated SVD $\mathrm{TF\mbox{-}IDF}(i,j)=n_{ij}\log\!\left(\frac{d}{b_i}\right),$9 downstream clustering and UMAP, often using dimensions $n_{ij}$0 because the first LSI component is correlated with sequencing depth and is excluded (Zandigohar et al., 2022). A plausible implication is that the triplet TF-IDF-like idea is not restricted to lexical terms: what matters is the existence of a local factor, a corpus-level factor, and optionally a third concentration or entropy-derived factor.

The significance-testing papers preserve the more classical bag-of-words representation. There the weight measures the extent to which a term’s occurrences are concentrated in a given document relative to the collection background, and the same logic is proposed for higher-order counted units such as triplets (Sheridan et al., 2020, Sheridan et al., 21 Jul 2025). The PLR burstiness paper goes further by turning the multi-factor statistic into a practical per-document weight $n_{ij}$1, later squashed by a sigmoid $n_{ij}$2 for classification experiments (Ahmed et al., 1 Apr 2026).

6. Empirical findings, limitations, and open cautions

The empirical record shows that tri-factor weighting can be effective, but it also shows that “better than TF-IDF” must be qualified. On 20 Newsgroups, the Term_Class Relevancy method reports roughly 86.88% to 89.57% accuracy for SVM (RBF) and roughly 90.38% to 93.01% accuracy for k-NN (k=10), and it outperforms the Bayes-posterior term-class relevance measure of Isa et al. 20. However, the same paper does not directly run a conventional TF-IDF baseline in the reported experiments, so its superiority over TF-IDF is structural and conceptual rather than established by a direct head-to-head table.

The troenpy-based TF-PI method is more directly comparative. On seven benchmark text datasets, the paper reports that TF-PI uniformly outperformed TF-IDF in kNN, with an average error reduction of 22.9% and the largest reported reduction of 53.4% on R8 (Zhang, 2023). That result is specific to a supervised setting, because PCF and ECIB require class labels and “cannot apply directly” when label information is unavailable.

The scATAC-seq study reports that TF-IDF transformation generally leads to improved clustering and biologically relevant feature extraction, and that among the tested transformations

$n_{ij}$3

performed best on the mouse forebrain dataset, while the entropy-based transformation underperformed in that setting (Zandigohar et al., 2022). The same source also cautions that there is still no universally agreed “gold transformation method,” especially across datasets.

The statistical-significance literature is deliberately more conservative. The hypergeometric-test study reports that TF-IDF performs comparably to the hypergeometric test on standard tasks rather than being clearly dominated by it (Sheridan et al., 2020). The Fisher exact test justification establishes a theoretical bridge from significance scores to TF-ICF and, under stronger assumptions, to TF-IDF, but its triplet generalization is presented as a reasoned extrapolation (Sheridan et al., 21 Jul 2025). The burstiness-based PLR study reports strong alignment with TF-IDF, including correlation $n_{ij}$4 on synthetic data and comparable document-classification accuracy on 20 Newsgroups and R8, but it also states that the method is approximation-based, requires hyperparameters $n_{ij}$5, and is more expensive than TF-IDF (Ahmed et al., 1 Apr 2026).

Several limitations recur across the literature. Supervised triplet-like schemes require labels (Guru et al., 2016, Zhang, 2023). Entropy- or burstiness-based variants add modeling choices and may underperform on some datasets (Zandigohar et al., 2022, Ahmed et al., 1 Apr 2026). Triplet-feature extensions are especially vulnerable to sparsity: counts such as $n_{ij}$6, $n_{ij}$7, and $n_{ij}$8 are often tiny, and the definition of the sample space $n_{ij}$9 depends on whether the triplet means a contiguous 3-gram, an unordered set, a phrase triplet, or another extracted unit (Sheridan et al., 21 Jul 2025). For that reason, the literature supports the phrase “triplet TF-IDF-like measure,” but not a single universally accepted formula.

The most precise synthesis is therefore twofold. In one sense, a triplet TF-IDF-like measure is a three-factor weighting scheme such as

$d_j$0

or

$d_j$1

In the other sense, it is a TF-IDF- or significance-based weighting on triplet features such as

$d_j$2

$d_j$3

or

$d_j$4

The literature does not collapse those senses into a single doctrine, but it consistently shows that TF-IDF-like weighting can be extended beyond the classical two-factor heuristic by incorporating class information, concentration, certainty, entropy, burstiness, or higher-order discrete features (Guru et al., 2016, Zhang, 2023, Zandigohar et al., 2022, Sheridan et al., 21 Jul 2025, Ahmed et al., 1 Apr 2026).