Semantic Similarity with Concept Senses: new Experiment
Bibliographic record
Abstract
This dataset represents the results of the experimentation of a method for evaluating semantic similarity between concepts in a taxonomy. The method is based on the information-theoretic approach and allows senses of concepts in a given context to be considered. Relevance of senses is calculated in terms of semantic relatedness with the compared concepts. In a previous work [9], the adopted semantic relatedness method was the one described in [10], while in this work we also adopted the ones described in [11], [12], [13], [14], [15], and [16]. We applied our proposal by extending 7 methods for computing semantic similarity in a taxonomy, selected from the literature. The methods considered in the experiment are referred to as R[2], W&P[3], L[4], J&C[5], P&S[6], A[7], and A&M[8] The experiment was run on the well-known Miller and Charles benchmark dataset [1] for assessing semantic similarity. The results are organized in seven folders, each with the results related to one of the above semantic relatedness methods. In each folder there is a set of files, each referring to one pair of the Miller and Charles dataset. In fact, for each pair of concepts, all the 28 pairs are considered as possible different contexts. REFERENCES [1] Miller G.A., Charles W.G. 1991. Contextual correlates of semantic similarity. Language and Cognitive Processes 6(1). [2] Resnik P. 1995. Using Information Content to Evaluate Semantic Similarity in a Taxonomy. Int. Joint Conf. on Artificial Intelligence, Montreal. [3] Wu Z., Palmer M. 1994. Verb semantics and lexical selection. 32nd Annual Meeting of the Associations for Computational Linguistics. [4] Lin D. 1998. An Information-Theoretic Definition of Similarity. Int. Conf. on Machine Learning. [5] Jiang J.J., Conrath D.W. 1997. Semantic Similarity Based on Corpus Statistics and Lexical Taxonomy. Inter. Conf. Research on Computational Linguistics. [6] Pirrò G. 2009. A Semantic Similarity Metric Combining Features and Intrinsic Information Content. Data Knowl. Eng, 68(11). [7] Adhikari A., Dutta B., Dutta A., Mondal D., Singh S. 2018. An intrinsic information content-based semantic similarity measure considering the disjoint common subsumers of concepts of an ontology. J. Assoc. Inf. Sci. Technol. 69(8). [8] Adhikari A., Singh S., Mondal D., Dutta B., Dutta A. 2016. A Novel Information Theoretic Framework for Finding Semantic Similarity in WordNet. CoRR, arXiv:1607.05422, abs/1607.05422. [9] Formica A., Taglino F. 2021. An Enriched Information-Theoretic Definition of Semantic Similarity in a Taxonomy. IEEE Access, vol. 9. [10] Information Content-based approach [Schuhmacher and Ponzetto, 2014]. [11] Linked Data Semantic Distance (LDSD) [Passant, 2010]. [12] Wikipedia Link-based Measure (WLM ) [Witten and Milne, 2008]; [13] Linked Open Data Description Overlap-based approach (LODDO) [Zhou et al. 2012] [14] Exclusivity-based [Hulpuş et al 2015] [15] ASRMP [El Vaigh et al. 2020] [16] LDSDGN [Piao and Breslin, 2016]
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.015 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.003 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.007 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".