Text Similarity in Vector Space Models: A Comparative Study

Shahmirzadi, Omid; Lugowski, Adam; Younge, Kenneth

doi:10.2139/ssrn.3259971

Cited by 19 publications

(6 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Each vector is determined via a term‐frequency inverse‐document‐frequency (TFIDF) approach that up‐weights rare words and down‐weights common words. Although the TFIDF approach is relatively simple, a benchmarking study on patent data shows that TFIDF performs well in situations of long, extended, and highly granular text (Shahmirzadi, Lugowski, and Younge, 2018). The similarity measure is then computed for each pair of patents by determining the angular distance between them using a cosine measure between the patents' two vectors.…”

Section: Results From Testing Assumptionsmentioning

confidence: 99%

Patent citations reexamined

Kuhn

Younge

Marco³

2020

The RAND J of Economics

135

View full text Add to dashboard Cite

Many studies rely on patent citations to measure intellectual heritage and impact. In this article, we show that the nature of patent citations has changed dramatically in recent years. Today, a small minority of patent applications are generating a large majority of patent citations, and the mean technological similarity between citing and cited patents has fallen considerably. We replicate several well‐known studies in industrial organization and innovation economics and demonstrate how generalized assumptions about the nature of patent citations have misled the field.

show abstract

Section: Results From Testing Assumptionsmentioning

confidence: 99%

Patent citations reexamined

Kuhn

Younge

Marco³

2020

The RAND J of Economics

135

View full text Add to dashboard Cite

show abstract

“…Consequently, many researchers primarily rely on exercise text modeling to characterize exercises [6][7][8]. For instance, Shahmirzadi et al [9] divide the topic text into sections, assign different weights to the resulting divisions, and combine each division vector to represent a complete topic. Zhang et al [10] exclude images and formula information from the input data, using a prefix tree-based lexicon-free word separation algorithm to extract features from the exercise text and reference answers.…”

Section: Unimodal Characterization Methodsmentioning

confidence: 99%

MIFM: Multimodal Information Fusion Model for Educational Exercises

Song,

Chen,

et al. 2023

Electronics

View full text Add to dashboard Cite

Educational exercises are crucial factors in the successful implementation of online education as they play a key role in assessing students’ learning and supporting teachers in instruction. These exercises encompass two primary types of data: text and images. However, existing methods for extracting exercise features only focus on the textual data, neglecting the rich semantic information present in the image data. Consequently, the exercise characterization vector generated by these methods struggles to fully characterize the exercise. To address these limitations, this paper proposes a multimodal information fusion-based exercise characterization model called MIFM. The MIFM model tackles the challenges of current exercise modeling methods by performing extraction and fusion operations on the heterogeneous features present in exercises. It employs a dual-stream architecture to separately extract features from images and text, and establishes connections between heterogeneous data using cross-modality attention methods. Finally, the heterogeneous features are fused using a Bi-LSTM combined with a multi-head attention mechanism. The resulting model produces a multimodal exercise characterization vector that fuses both modalities and incorporates three knowledge elements. In the experiments, by using the model to replace the exercise characterization modules in the three educational tasks, specifically, it achieves an increased ACC value of 72.35% in the knowledge mapping task, a heightened PCC value of 46.83% in the exercise difficulty prediction task, and an elevated AUC value of 62.57% in the student performance prediction task.

show abstract

“…Patent metinlerinin benzerliğinin hesaplanmasında TF-TBF yöntemi nitelikli sonuçlar ortaya koymaktadır [18]. Klasik TF-TBF yönteminden başka TF-TBF yöntemlerinin bazı varyantları da patentpatent benzerlik analizinde kullanılmaktadır [18].…”

Section: )unclassified

A Review on the Determination of Semantic Similarity of Patent Documents

KAYAKÖKÜ,

TÜFEKCİ

2023

Journal of Polytechnic

View full text Add to dashboard Cite

Teknik anlamda en güncel bilgileri barındıran, yüksek hacmiyle bilgi keşfi açısından müthiş bir potansiyele sahip olan ve teknoloji yönetimi alanında kilit bir rol üstlenen patent verisinin işlenmesinde patent madenciliği çalışmaları giderek önem kazanmaktadır. Patent verisi içerisinde bulunan yapısal veya yapısal olmayan verilerin hepsi önemli olsa da, patent madenciliği çalışmalarının en kritik hedefi patent dokümanlarının anlamsal benzerliğini tespit edebilmektir. Patentlerin anlamsal benzerlik tespiti ile patent başvuru sürecinin en zor ve en çok vakit alan safhası olan patentlenebilirlik kriterlerinin tespitinin otomatik olarak yapılabilmesi mümkün olacaktır. Patent metinlerinin, metin madenciliği yöntemleri ile yapısal hale getirilerek birbirine ne kadar benzediklerini tespit etmek için küme teorisi yaklaşımları, vektör uzay modeli yaklaşımları veya ontoloji vb. bilgi kaynaklarından faydalanılan yaklaşımlar mevcuttur. Ancak patent metinlerinin karmaşık yapısı ve kendine has terminolojisi sebebiyle bu yöntemlerden hedeflenen verim alınamamaktadır. Bu eksikliği gidermek için kullanıldığı her alanda büyük başarılar ortaya koyan derin öğrenme yöntemlerinden, patent metinlerinin anlamsal olarak karşılaştırılmasında da faydalanılması gerekmektedir. Bu alanda çalışmalar yapılmasına rağmen etkin bir şekilde patentlenebilirlik tespiti yapabilen modeller henüz başlangıç aşamasındadır. Nitelikli bir model geliştirilerek patentlenebilirlik tespiti yapıldıktan sonra patent araştırma raporunun otomatik olarak hazırlanması teknoloji yönetimi alanındaki büyük ihtiyacın karşılanabilmesi adına önemli bir adım olacaktır.

show abstract

Text Similarity in Vector Space Models: A Comparative Study

Cited by 19 publications

References 13 publications

Patent citations reexamined

Patent citations reexamined

MIFM: Multimodal Information Fusion Model for Educational Exercises

A Review on the Determination of Semantic Similarity of Patent Documents

Contact Info

Product

Resources

About