Build a semantic similarity network

Constructs an undirected k-nearest-neighbour network based on work embeddings. Each work is connected to its k nearest neighbours in embedding space (cosine similarity). The edge weight is the cosine similarity between the connected works.

Usage

sm_network_semantic(corpus, k = 10L, call = rlang::caller_env())

Arguments

corpus

An sm_corpus object. Must contain an embeddings matrix (see sm_embed_works()).

k

Integer; number of nearest neighbours per work. Defaults to 10L.

call

Caller environment for error reporting.

Value

A tidygraph::tbl_graph object (undirected). Nodes carry name (work ID) and columns from corpus$works. Edges carry a weight column (cosine similarity, between 0 and 1).

Details

Embeddings must be present in corpus$embeddings (a numeric matrix with row names matching work_id). Use sm_embed_works() to compute them.

The function computes cosine similarity via matrix multiplication on L2-normalised vectors and selects the top-k neighbours for each work.

Empty input or missing embeddings returns an empty undirected tbl_graph.

See also

Other networks: sm_network_citation(), sm_network_cocitation(), sm_network_collab(), sm_network_coupling(), sm_network_coword()

Examples

corpus <- sm_example_corpus(with_embeddings = TRUE)
g <- sm_network_semantic(corpus, k = 5L)
g
#> # A tbl_graph: 200 nodes and 768 edges
#> #
#> # A bipartite simple graph with 5 components
#> #
#> # Node Data: 200 × 16 (active)
#>    name      doi   title abstract  year type  source_id cited_by_count oa_status
#>    <chr>     <chr> <chr> <chr>    <int> <chr> <chr>              <int> <chr>    
#>  1 W0000000… 10.1… Tumo… This st…  2023 jour… S0000000…              3 green    
#>  2 W0000000… 10.1… Biom… This st…  2020 jour… S0000000…              9 hybrid   
#>  3 W0000000… 10.1… Colo… This st…  2024 jour… S0000000…             28 green    
#>  4 W0000000… 10.1… Tumo… This st…  2020 revi… S0000000…             29 green    
#>  5 W0000000… 10.1… Clin… This st…  2020 jour… S0000000…             16 hybrid   
#>  6 W0000000… 10.1… Tumo… This st…  2018 jour… S0000000…             16 closed   
#>  7 W0000000… 10.1… Sing… This st…  2021 lett… S0000000…             26 bronze   
#>  8 W0000000… 10.1… Colo… This st…  2020 jour… S0000000…              7 hybrid   
#>  9 W0000000… 10.1… Gene… This st…  2024 revi… S0000000…              6 gold     
#> 10 W0000000… 10.1… Colo… This st…  2019 revi… S0000000…              5 gold     
#> # ℹ 190 more rows
#> # ℹ 7 more variables: language <chr>, pmid <chr>, arxiv_id <chr>,
#> #   openalex_id <chr>, is_retracted <lgl>, retraction_date <date>,
#> #   last_refreshed <dttm>
#> #
#> # Edge Data: 768 × 3
#>    from    to weight
#>   <int> <int>  <dbl>
#> 1     1    74  0.928
#> 2     1    92  0.926
#> 3     1   101  0.932
#> # ℹ 765 more rows